Police AI Deployment Tracker

Deployment

Legal basis & oversight

Legal basis

Varies by state; several states and 20+ cities have banned or restricted police facial recognition

Oversight

Fragmented. State and city-level restrictions in some jurisdictions, but no single federal framework.

What we know

Clearview markets its database of images scraped from the web as an investigative lead generation tool, and states that results are indicative rather than definitive matches requiring further verification. The roughly 3,100 US agency figure originates from the company itself and should be read as a historical vendor estimate, not a verified current count. A caution on attribution that matters for accuracy: not every US facial recognition wrongful arrest is a Clearview case, and this entry should not absorb them. The documented Clearview-attributed cases are Randal Quran Reid (Georgia, arrested for a Louisiana offence he had no connection to) and a Jefferson Parish Sheriff's Office case that led to a $200,000 settlement. By contrast, the June 2026 Florida suit filed by the ACLU on behalf of Robert Dillon concerns Pinellas County's FACES system, not Clearview, and the ACLU's April 2026 update identifying a fourteenth publicly known US wrongful arrest does not attribute a vendor in most cases. Legal position: the Seventh Circuit overturned the class settlement in the Illinois BIPA consumer privacy litigation on 13 July 2026. Federal contracting continues. ICE Homeland Security Investigations signed a $9.2m contract in September 2025; US Customs and Border Protection signed a one-year $225,000 contract on 11 February 2026 for tactical targeting operations; and the US Army's 1st Special Forces Command has a contract with a base period running March 2026 to March 2027, with options to 2030 and a total potential value of $339,415.

Source

Company statements; ACLU documentation of wrongful arrest cases; contract reporting · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Ordinary municipal procurement; no dedicated statute

Oversight

None identified externally; Antioch's staff report frames dispatchers as retaining prioritisation and dispatch decisions

What we know

Assistive AI layer for 911 call centres: transcribes calls live, translates in real time via synthetic voice, and extracts addresses/incident details automatically. Reports 1,000+ agency partnerships covering ~100 million people; combined with Carbyne, Axon states the pair process 80m+ 911 calls/year. Trade publication Officer.com reports (Aug 2026, citing Axon's 911-division CEO) that in Fort Worth TX the platform handles non-emergency calls fully autonomously, thousands per month with no human call-taker -- a categorical step beyond every other control-room entry on this tracker, which all describe AI assisting a human who retains the decision. This specific claim is vendor-sourced (a company CEO in trade press, not a Fort Worth city source) and is recorded with that caveat; if independently confirmed it would be the first fully automated call-handling deployment found in this research. Contrasts directly with this site's existing UK 101-triage entry, which records the UK government's position that it 'does not support police use of AI to answer 999 calls'. Confirmed municipal procurement: Antioch CA (March 2026 sole-source staff report) and Alameda County Sheriff's Office.

Source

Officer.com; Antioch City Council staff report via Contra Costa News · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Varies by state; no dedicated federal ALPR statute

Oversight

Fragmented. State-level audits in some cases (e.g. Illinois), but no single national body.

What we know

Fragmented by design: thousands of individual agencies each contract with Flock separately, with far less national standardisation of retention or access than a single national system would have. Scale as of August 2026 is roughly 120,000 cameras across 49 states, with a company-cited reach of 6,000 to 7,000 law enforcement agencies; sources differ on the exact headcount and the figures move quickly. That fragmentation is where the most serious documented problems have surfaced. An Illinois state audit found federal immigration authorities had gained access to state plate-read data through an undisclosed CBP arrangement, in apparent tension with state law, and 47 agencies were subsequently removed. A Dayton, Ohio audit in May 2026 found camera data searched more than 7,100 times for immigration enforcement in breach of city policy; the city covered all 72 cameras with bin bags pending contract exit. In San Francisco, 299 improper queries were made over about a year by a regional intelligence centre. The University of Washington Center for Human Rights documented federal access to Washington state networks through front-door, back-door and side-door routes. The Electronic Frontier Foundation reported over eighty agencies entering search justifications using language perpetuating stereotypes about Romani people. Individual officer misuse is now documented repeatedly: a Washington Post investigation found dozens of officers using the network to surveil former partners and romantic interests, and separate 2026 cases include a Georgia officer arrested in August, an Arizona officer who resigned after allegedly tracking his wife, and a North Carolina officer charged over 31 searches of her boyfriend's ex-wife's vehicle. Vendor response, 13 August 2026: default retention cut from 30 days to 7; the previously optional abnormal-search monitoring tool made universal, having been enabled by only about a third of customers; automatic lockouts for suspicious search patterns; case numbers required for extended retention. Flock had earlier removed federal agencies from its national search network (August 2025) and required a NIBRS offence type on every search (December 2025). Contraction is substantial but so is growth: around 53 cities had cancelled by late June 2026 and over 20 jurisdictions moved to drop Flock in July 2026 alone, while roughly 800 US cities approved new contracts during 2026. Ring cancelled a planned Flock integration in March 2026. Washington State's Driver Privacy Act (ESSB 6002), signed 30 March 2026 and effective immediately, imposes a 21-day retention cap, permitted-purpose limits, bans on immigration and reproductive-healthcare and protected-activity uses, collection bans near sensitive sites, mandatory registration with the Attorney General and annual audits.

Source

Company statements; Illinois state audit findings, reported by State of Surveillance and the ACLU · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

DPA 2018 Part 3; ANPR national standards

Oversight

National ANPR Service; Biometrics and Surveillance Camera Commissioner

What we know

The National ANPR Service processes tens of millions of reads daily. Retention periods and access controls are set nationally.

Source

Home Office National ANPR Service documentation · last checked 2026-08-18

Deployment

Legal basis & oversight

Legal basis

Home Office programme with a published privacy notice; no dedicated statute identified

Oversight

None named externally; Home Office CAID team

What we know

CAID is the UK's national hash-matching database of child sexual abuse imagery, live since Dec 2014 and used by all territorial forces, Police Scotland, PSNI and the NCA. The AI layer -- the Vigil AI CAID Classifier, built with Roke -- trains on the CAID corpus to detect previously unseen CSAM, including deepfake CSAM, and grade severity against government categories, going beyond hash-matching's ability to only catch already-known material. Roke describes it as (in the supplier's own words) the first operationally viable classifier able to both detect and grade. Home Office had invested £18.2m by 2019 with a further £1.76m on AI/fast-forensic tools. A May 2024 Home Office market notice signalled a possible move to public cloud hosting, reversing a 2020 legal opinion against cloud hosting for this data; outcome not established. This is the only documented operational police deployment of AI against synthetic/deepfake media found in this research pass.

Source

GOV.UK CAID privacy notice; Roke white paper; Computer Weekly · last checked 2026-08-24

Deployment

Legal basis & oversight

Legal basis

Internal force policy; no dedicated statutory framework for generative AI use

Oversight

Individual force AI-use policy; no independent national oversight body confirmed

What we know

At least 21 English forces were using Microsoft 365 Copilot as of February 2026, on a Sky News survey; no later recount has been published, so that figure should be read as a point-in-time count rather than a current one. Governance moved during 2026. Alex Murray, head of the new PoliceAI centre, told the Financial Times in June 2026 that he had instructed several forces in England and Wales to stop using commercially available AI to turn interview recordings into court-ready statements, and warned forces not to use AI for disclosure schedules until compliance checks were complete. Specific forces were not named. The Home Office subsequently listed exploring Copilot capability and establishing appropriate governance and safeguards among PoliceAI's first-year priorities. Individual misuse cases are now emerging, which is a different failure mode from the West Midlands hallucination and worth recording separately. A Derbyshire detective chief inspector is under IOPC-led independent investigation for alleged gross misconduct over using AI to maintain investigative decision logs; a second Derbyshire officer faces a criminal investigation for allegedly using AI to create evidential material in a number of cases; and a Gloucestershire Police staff member was dismissed for repeatedly bypassing blocks to use an AI chatbot for police work.

Source

Sky News investigation, reported via Resultsense · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Not established in available sources

Oversight

Subject to legal challenge; independent review conducted by Eticas

What we know

At least 23 predictive models developed, with 13 risk models carrying performance data used between 2017 and 2024, covering missing people, antisocial behaviour, and likelihood of committing or suffering crime. An Eticas review of more than 36,000 disclosed model performance scores found low precision, including a burglary prediction model below 10 percent precision for over three years. The force's Offender Management App, designed to hold data on around 300,000 people and described internally as a league table of the area's most dangerous criminals, correctly predicted only one in three actual offenders while flagging one in four non-offenders on July 2025 audit data. A subject is pursuing legal challenge.

Source

Liberty Investigates and WIRED; independent review by Eticas · last checked 2026-08-20

Deployment

Legal basis & oversight

Oversight

A Memorandum of Understanding between the Bedfordshire Police and Crime Commissioner and the Chief Constable governs use, and an ethics panel assesses compliance and community impact. Both are stated on the force's own live facial recognition page and in its news release of 11 September 2025. Neither the Memorandum of Understanding nor any ethics panel minutes were read in this run.

What we know

Established by two Bedfordshire Police publications read on 9 September 2026: the force's live facial recognition page, and its news release "Police announce use of Live Facial Recognition across Bedfordshire", published 11 September 2025. The release describes the roll out of a new capability rather than a trial, and gives the first deployment as Friday 19 September 2025 in Bedford town centre, following earlier use at the Bedford River Festival. Superintendent Ian Taylor is named as the force's LFR strategic lead and Police and Crime Commissioner John Tizard states that he agreed a Memorandum of Understanding with the Chief Constable. The current force page publishes a table of planned deployments in Bedford and Luton town centres through September 2026 and states that the force is "now at a stage where we are confident in its ability to accurately identify missing people or people of interest". That continuing published deployment schedule, and the force's own framing of the September 2025 launch as a roll out rather than a trial, are the basis for recording the status as operational. Legal basis is deliberately left empty: the force page cites the Data Protection Act 2018, MOPI and the Criminal Procedures and Investigations Act 1996 in relation to retention of data, and section 60AA of the Criminal Justice and Public Order Act 1994 in relation to face coverings, but names no statutory power for the deployment itself. Vendor is deliberately left empty: neither force publication names a supplier. A supplier is named for Bedfordshire in NEC Software Solutions marketing material, which is vendor-sourced and has not been treated as establishing anything. Unlike Essex, Surrey and Northamptonshire, the force publishes forthcoming deployments only and carries no historical register of past deployments with watchlist sizes, alerts or outcomes, so no arrest or alert figures are recorded here.

Source

Bedfordshire Police live facial recognition page and force news release · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

EU Regulation establishing the Entry/Exit System

Oversight

European Commission; national data protection authorities and the European Data Protection Supervisor

What we know

On a non-EU traveller's first entry and exit the system electronically registers passport data plus biometrics, specifically a facial image and fingerprints, with entry and exit records. On subsequent crossings it performs fast biometric verification. It replaces manual passport stamping and covers short stays of 90 days in any 180. Became operational on 12 October 2025 with a six-month progressive rollout, and fully operational from 10 April 2026 at all external Schengen crossing points except Cyprus and Ireland. This is among the largest-scale biometric systems in this catalogue, and differs from investigative facial recognition in that it verifies an asserted identity against government-held documents rather than searching to name an unknown person.

Source

European Commission · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Not detailed in the source reviewed

Oversight

Not detailed in the source reviewed

What we know

The Brussels police zone's own description states it uses BriefCam to convert hours of CCTV footage into minutes of targeted review. Recorded with lower confidence than the French and Dutch entries above: the source is the police force's own promotional description rather than an independent inspection or an official transparency register, and no usage statistics or legal framework were available in the material reviewed. Sourcing note, verified August 2026: the available documentation is vendor case-study material rather than a police or government disclosure, and should be read as such. It describes the deployment as operated by the Brussels Regional Informatics Centre (BRIC), which since around 2014-2015 has centralised video from six local police zones, the Metro system and Bruxelles Mobilite onto a single platform. In that documentation BriefCam functions as an analytics extension within a Genetec Security Center video management deployment, rather than as a standalone force procurement. Two caveats follow: the operator is a regional IT body serving multiple police zones rather than a single police zone, and the vendor material does not establish the specific start date recorded on this entry. Both points would benefit from a Belgian procurement or oversight-body source.

Source

BriefCam vendor case study; Genetec customer story · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Police have relied on section 75 of the Sicherheitspolizeigesetz. In a decision reported on 26 July 2026 the Datenschutzbehoerde found that provision insufficient to authorise automated biometric comparison, holding that there is currently no adequate legal basis for it. The Interior Ministry said it would appeal to the Bundesverwaltungsgericht, and the decision was not final when reported.

Oversight

Datenschutzbehoerde, the Austrian data protection authority, which upheld a complaint brought by the digital rights organisation epicenter.works on behalf of an activist photographed at a climate protest in Vienna in March 2023 and later matched using facial recognition. Parliamentary scrutiny has run through written questions to the Interior Minister.

What we know

The deployment itself is well established. ORF, the Austrian public broadcaster, reported on 8 September 2020 that the system passed from roughly a year of testing into regular operation at the start of August 2020, that it had been used 581 times and had identified 83 suspects by that point, that it cost 450,000 euros, and that it compares images against an Interior Ministry photograph database of suspects. Netzpolitik.org, 16 September 2020, gives the vendor as Cognitec Systems of Dresden, the software cost as 448,813 euros, the database as the police Lichtbilddatei of photographs from earlier criminal proceedings, the same figure of 581 searches, and reports that the system was used on images from the summer 2020 protests in the Favoriten district of Vienna, where 47 known persons were identified. The two cost figures are the same number reported gross and net and are not treated as a conflict. All usage figures date from 2020 and no current figures were found; they are recorded here rather than in a field for that reason. The legal finding is the significant recent development: reporting by the Austrian news agency APA on 26 July 2026, carried by Salzburger Nachrichten and others, and by epicenter.works on 23 July 2026, establishes that the Datenschutzbehoerde found the use of automated facial recognition on protest photographs unlawful for want of a sufficient legal basis. The status remains operational because nothing read establishes that use has stopped - the Interior Ministry publicly defended the system and announced an appeal. No parliamentary answer was read directly in this run: the Austrian Parliament's page for answer 6962/AB confirms the answer was given by Interior Minister Karl Nehammer on 16 August 2021 but the substance sits in a linked PDF that was not retrieved.

Source

ORF · last checked 2026-09-06

Deployment

Legal basis & oversight

Legal basis

Departmental policy; no statutory basis or external regulator identified

Oversight

Internal supervisory review; use-of-force and problematic language flags require sergeant review

What we know

Implemented on 1 February 2024 for half the patrol force as a randomised controlled trial, which is why this deployment rests on peer-reviewed evidence rather than vendor material. Natural language processing runs over the audio of every body-worn camera activation once cameras are docked, producing transcripts and automatic flags for professionalism, de-escalation, use of force and language. Apache Junction Police Department has also been studied. Transcription and annotation accuracy is reported to exceed 90 percent, a vendor-derived figure cited within the peer-reviewed study.

Source

Policing: A Journal of Policy and Practice (Oxford University Press), peer-reviewed · last checked 2026-08-20

Deployment

What we know

Established by two independent South African publications, both quoting the City's Mayoral Committee Member for Safety and Security, Alderman JP Smith, by name. Engineering News, carried by Polity on 28 July 2023: the City already had fixed ANPR cameras, had piloted dashboard-mounted units the previous year, and would fit the technology to more enforcement vehicles in the 2023/24 financial year. The cameras scan plates to detect outstanding fines, warrants, cloned plates and stolen vehicles. Budget of 26 million rand covering ANPR, body cameras, dashcams and digital evidence management. ITWeb, 19 November 2024: 290 in-vehicle ANPR cameras deployed during the 2023/24 financial year, operated with round-the-clock monitoring by the Metro Police Strategic Surveillance Unit. Reported results include one motorist arrested carrying 262 fines and 114 warrants totalling 209,200 rand, and 360 arrests with 71,843 traffic and by-law offences recorded in a single week. Smith is quoted: 'This system is improving efficiency by allowing our officers to focus on the right vehicles, which means it is increasingly more difficult to hide.' Note that this is a municipal police service, not the South African Police Service. The two bodies are distinct and a future pass should keep them separate. Nothing read in this run establishes any SAPS national ANPR programme. Fields left out and why. 'since' is omitted: fixed cameras predate the 2023 article by an unstated period and the 290 in-vehicle units date from 2023/24, so no single start date is established. 'vendor' is omitted: no supplier is named in either article. 'legalBasis' and 'oversight' are omitted: nothing specific was established.

Source

ITWeb · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

Not established in available sources

Oversight

Not established in available sources

What we know

The replacement for Action Fraud and the National Fraud Intelligence Bureau, delivered as a transformation programme of around 150 million pounds and released in stages, with an interim release in November 2024 and a second in March 2025, with staff using the new platform. City of London Police originally stated that the replacement service would incorporate AI and machine learning to improve proactive fraud and cybercrime prevention and intelligence analysis. Recorded deliberately as a platform incorporating AI capabilities rather than as an AI system. It is primarily a national fraud and cyber reporting and intelligence platform, and precisely which AI models are currently operational within it is not established. Update, verified August 2026: the service has since launched publicly under the brand name Report Fraud, going live on 4 December 2025 and fully replacing Action Fraud, with a public awareness campaign from January 2026. Behind it, the Report Fraud National Crime Analysis Service (N-CAS) replaces Action Fraud's back-end analysis system serving the National Fraud Intelligence Bureau. Suppliers are Capita (contact centre and public-facing reporting technology) and PwC UK (crime and intelligence management technology underpinning the NFIB), appointed June 2023. Note a discrepancy in published cost figures worth recording: the force's own statements cite a £150m investment, while an earlier City of London Police budget reported by Recorded Future News put the estimated total cost of FCCRAS at just over £31m. The two figures may cover different scopes; neither has been reconciled publicly.

Source

City of London Police, own statements and annual report · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No dedicated statute for the tools themselves. The disclosure was published under Section 5(3) of Act No. 106/1999 Coll. on free access to information. The force states it is preparing for obligations under EU Regulation 2024/1689 (the AI Act).

Oversight

No dedicated regulator identified. The force states it plans to carry out a fundamental-rights impact assessment once AI Office and European Commission methodology is published.

What we know

The Czech Police's own public statement is unusual in explicitly distinguishing which of its forensic tools it considers AI/machine learning (this list) from which it does not: it names TrueAllele, STRmix, LRmix, StatistiX, CODIS and FamLinkX as fixed statistical or probabilistic systems it does not classify as AI. The force states the AI/ML tools listed here are used for preliminary information processing and do not independently collect, evaluate or generate evidence. This entry bundles several distinct tools under one entry because the source describes them collectively rather than confirming each one's individual deployment scale. Sourcing, verified August 2026: this comes from the Czech Police's own freedom-of-information response, published on its website under Section 5(3) of Act No. 106/1999 Coll. The wording confirms the entry's central point directly. The force states it does not currently use any AI system to independently collect, evaluate or generate evidence, and that tools applying machine learning technologies are used for partial tasks in the preliminary processing of information, naming ABBYY FineReader, BriefCam, Magnet AXIOM, Cellebrite UFED, speech-to-text from Phonexia and Newton, together with outputs from Ministry of the Interior security research projects. The same response records that the force plans to conduct a fundamental-rights impact assessment under the EU AI Act, but that it does not yet know the required content, scope or form because the relevant AI Office and European Commission methodology has not been published. For independent context rather than force self-description, the civil society organisation Iuridicum Remedium has published a mapping study of Czech police AI use ('Cesky Minority Report'), which covers automated recognition of faces, sounds and text and is a useful counterweight to a disclosure written by the force itself.

Source

Policie CR, published freedom-of-information response · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No dedicated statute; matches against biometrics held under the Criminal Procedure (Identification) Act 2022

Oversight

None identified; constitutionality is contested before the Supreme Court of India

What we know

Matches faces from cameras/video one-to-many against Delhi Police and National Crime Records Bureau databases, treating an 80%-similarity score as a positive match. Procured after a 2018 Delhi High Court direction, originally for tracing missing children; since used for crowd screening (2019 political rally), the 2020 Delhi riots, 2021 Red Fort clashes and Jahangirpuri riots. FR vans were deployed at student protests from 20 July 2026; a counter-affidavit sworn by a Deputy Commissioner and filed before the Supreme Court (Shailendra Mani Tripathi v Union of India, WP 280/2026) states the system generated matches for 2,873 individuals between 20-26 July, 92 linked to 10+ cases. No judgment yet -- an affidavit in live proceedings, not a ruling. Delhi Police told the Delhi HC in 2018 its own system's accuracy was 2%; the Internet Freedom Foundation has established via RTI that there is no rule governing use, no privacy impact assessment and no retention policy. India has no dedicated law governing biometric surveillance.

Source

Bar and Bench court reporting; MediaNama RTI findings; Internet Freedom Foundation · last checked 2026-08-24

Deployment

What we know

Oyoon — Arabic for 'eyes' — pools camera feeds from tourist sites, public transport, the road network and public buildings into a Dubai Police command centre, where software matches faces against police records and reads vehicle registration plates to flag wanted people and vehicles. Dubai Police run it with partner bodies including the State Security Agency, Dubai Municipality and the Roads and Transport Authority. Figures, each with its source. In March 2019 Gulf News quoted First Lieutenant Ali Al Shehi of the Dubai Police Criminal Investigation Department saying the cameras 'helped arrest 319 suspects in different areas in Dubai through their automatic facial recognition technology' in the preceding year, and gave 5,000 cameras in the Metro system alone. In July 2021 Gulf News reported Sheikh Mohammed bin Rashid Al Maktoum, briefed on the programme alongside the Dubai Police Commander-in-Chief, saying Dubai is monitored by more than 300,000 cameras, and reported emergency response times falling from 6.46 minutes to under three minutes. The Oxford Institute of Technology and Justice, in a country study current to March 2026, describes Oyoon as using facial recognition and behaviour analysis to identify suspects and track vehicles and plates across more than 5,000 cameras at tourist sites, public transport and traffic. Legal position. The Oxford study states that there is no dedicated legislation in the UAE governing the use of AI in criminal proceedings, and that the framework rests on general instruments — the Charter for the Development and Use of AI (2024), the AI Ethics Principles and Guidelines (2022) and the AI Adoption Guideline in Government Services (2023). 'legalBasis' and 'oversight' are therefore omitted rather than filled: the honest statement is that neither is established. Weaknesses, stated plainly. No Dubai Police publication was read directly in this run; dubaipolice.gov.ae could not be reached. Everything above is official statement relayed by reputable regional journalism, plus one academic source. The camera counts are not consistent with one another and are not treated as a single figure: 5,000 is the Metro estate, 300,000 is the ruler's figure for the emirate as a whole, and the two are reported five years apart. No independent accuracy or false-match figure exists for the facial recognition component, and none is asserted here.

Source

Gulf News · last checked 2026-09-07

Deployment

Legal basis & oversight

Legal basis

Not detailed in the source reviewed

Oversight

Documented via the Dutch government's own public algorithm register, a transparency mechanism

What we know

The Dutch police's own entry in the government's official algorithm register describes a tool using convolutional neural network and deep learning models to identify moving objects in video and classify them by characteristics including type (person or vehicle), colour, size and direction of movement, letting investigators combine filters such as a person on a bicycle wearing a particular colour travelling in a particular direction. Results are reviewed by police staff; the register states the algorithm itself does not make investigative decisions. This is one of the clearest official public descriptions found anywhere of how this class of video analytics tool actually works.

Source

Dutch government algorithm register (Algoritmeregister) · last checked 2026-08-26

Deployment

What we know

Established by two sources read in this run, one of them an official state government publication. Enugu State Government, 3 July 2026, reporting the Inspector-General of Police Tunji Disu's visit of 28 June 2026: the Command and Control Centre uses AI-enabled cameras with long-range zoom, operates real-time monitoring across the state with 'officers dedicated to certain areas of the state watching, seeing what is happening', and is linked to the Distress Response Squad, described as a special police unit with over 150 security vehicles fitted with AI-enabled cameras and connected to the centre. Disu is quoted: 'There is no place that is left ungoverned from the control room.' Economy Post, 8 September 2026, an investigation: over 1,000 cameras installed across roads, markets, junctions and public facilities; facial recognition and automatic number plate recognition; drones with thermal and zoom capability for areas without fixed coverage; AI analytics designed to detect accidents, wrong-way driving, loitering and unusual gatherings; a command centre at the Enugu State Government House operating 24 hours a day with real-time coordination with security agencies; and budget allocations of about 4.657 billion naira in 2024 and 10 billion naira in 2025 for CCTV, drones and facial recognition. The same investigation reports that no publicly available information identifies the companies that supplied or installed the equipment. What is and is not established. The AI-enabled camera network, the command centre, the 2024 launch and the integration with a named police unit rest on both sources including the state government's own page, and are treated as established. The facial recognition and number plate recognition capabilities, the camera count and the budget figures rest on the Economy Post investigation alone and are recorded here as attributed reporting rather than as confirmed capability - which is why the technology field describes an AI-enabled camera network rather than naming facial recognition. Legal position. The Economy Post investigation reports that Enugu State has no dedicated safe city or surveillance law, that the state Attorney-General confirmed draft legislation is still being worked on, and that only the Nigeria Data Protection Act 2023 currently applies. 'legalBasis' is left empty rather than filled with that, because an absence of a specific legal basis is a finding about the deployment and not a legal basis; it is stated here in the notes instead. Fields left out and why. 'vendor' is omitted - the investigation specifically records that no supplier is publicly identified. 'oversight' is omitted: none established. This is a state government facility rather than a federal police programme. Nothing read in this run establishes any comparable AI deployment by the Nigeria Police Force nationally.

Source

Enugu State Government · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

Not specified in source; subject to an ICO compliance audit

Oversight

ICO (Information Commissioner's Office) -- audited as part of a five-force programme

What we know

Essex Police has used live facial recognition since summer 2024, supplied by Corsight AI (Apollo 4) with Digital Barriers. The force paused deployments in March 2026 after two independent evaluations it had itself commissioned, under its Public Sector Equality Duty, returned conflicting findings. The University of Cambridge study, published 12 March 2026, ran a controlled field experiment with 188 volunteers during a live deployment in Chelmsford. At the operational threshold in use, the system correctly identified roughly half of watchlist subjects who passed the cameras. It found the system more likely to correctly identify men than women, and statistically significantly more likely to correctly identify Black participants than participants of other ethnicities. The direction of that finding is worth stating precisely, because it is the opposite of the usual concern: the disparity was in the CORRECT identification rate, not the false positive rate. Of six false positives observed, four involved Black individuals, but the researchers described this as suggestive rather than conclusive given the small number of events. Separate National Physical Laboratory testing of Corsight Apollo 4, also published March 2026, found a true positive rate of 89% and a false positive rate of 0.017% against an 18,000-image watchlist, and reported no statistically significant demographic bias. The two studies do not agree, and this entry records that rather than picking one. A force impact assessment covering more than 40 deployments between August 2024 and February 2025 reported approximately 1.3 million faces scanned, 123 interventions, 48 arrests and one confirmed mistaken intervention. Essex Police said it decided to pause deployments while working with the algorithm provider to review results and update the software. The pause was disclosed through the ICO’s audit of the force. Deployments have since resumed. The force’s own live facial recognition page, read on 3 September 2026, states that Essex Police “have been deploying Live Facial Recognition (LFR) technology since summer 2024”, carries no reference to a pause, and publishes deployment results running to 16 August 2026 — most recently at Tendring, where 6,061 faces were scanned, producing one positive alert, one intervention and no incorrect alerts. The date deployments restarted is not established. The force page presents an unbroken account of use since 2024 rather than recording the suspension and its end, so the resumption can be dated only to somewhere between March and August 2026. Secondary reporting gives figures for the period after resumption but was not checked against a force publication and is not relied on here. The March 2026 pause and the conflicting evaluations above are retained deliberately. They are the most consequential facts on this record and remain true as history.

Source

Essex Police live facial recognition page; BBC News; Computer Weekly; Cambridge evaluation; NPL; ICO audit · last checked 2026-09-03

Deployment

Legal basis & oversight

Legal basis

No dedicated statute; US courts have given wide scope to surveil prisoner calls for safety/crime prevention; all parties are notified calls are monitored

Oversight

None identified externally

What we know

Transcribes prisoner telephone calls at scale using Amazon Transcribe plus keyword and semantic analysis, flagging calls for investigators. Now extended (Aug 2026) into continuous voice-identity tracking, which alerts when a caller's voice does not match the person assigned to the PIN used, and into AI video analysis for facility monitoring. Florida DOC paid $2.5m from Aug 2023 for up to 50 million minutes/year; Georgia DOC monitored 7.5m+ calls by April 2021, flagging 1,612 as showing evidence of criminal activity; company reports 65m+ calls/year across seven states. A Thomson Reuters Foundation investigation found the system used beyond its stated gang/violence rationale, including flagging a call reporting a COVID outbreak and one about unemployment benefits, and reportedly configured to flag Spanish words for 'lawyer' -- raising a legal-privilege concern. Differs from the existing seized-device analysis entry in this category by monitoring live telephone infrastructure continuously and at population scale, including family members who have not been convicted of anything.

Source

Thomson Reuters Foundation investigation; Biometric Update · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Extended to 31 December 2030 by Article 19 of the loi Ripost, definitively adopted by Parliament on 21 July 2026, for the sole purpose of preventing terrorism and serious threats to personal safety. Biometric identification remains excluded. Real-time facial recognition itself remains prohibited in French public space.

Oversight

Conseil constitutionnel; CNIL (data protection authority)

What we know

Piloted at the Paris 2024 Olympics for crowd management and threat detection, flagging anomalies for a trained human operator rather than acting automatically. The legislative history is unusually tangled and worth recording precisely, because the experiment has now been extended twice after being struck down once. The original Olympic-law authorisation was due to expire 31 March 2025. A first extension attempt, attached to transport safety legislation, was censured by the Conseil constitutionnel on 24 April 2025. The experiment was then extended to 31 December 2027 by the JOP 2030 law, upheld by the Conseil constitutionnel in March 2026, with an evaluation report due 30 September 2027. The loi Ripost, definitively adopted 21 July 2026, extended it again to 31 December 2030 and widened its scope from public roads and event perimeters to the interior of buildings open to the public, subject to Interior Ministry authorisation. Ripost also opens a separate and distinct experiment, running to 31 December 2027, permitting algorithmic analysis of surveillance footage in retail premises, supermarkets and shopping centres for the sole purpose of preventing theft. That is framed with its own safeguards: no biometric identification or facial recognition, mandatory human review, a CNIL impact assessment, a register of actions taken on alerts, and a published conformity attestation. Worth noting against the direction of travel: the official evaluation committee found the Olympic pilot's technical performance uneven and its operational value limited. The extensions have proceeded regardless.

Source

French government legislation; Conseil constitutionnel rulings; Amnesty International France and Statewatch reporting · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Contested. The Code of Criminal Procedure restricts real-time TAJ checks to specially authorised officers during an investigation; independent reporting finds the database used for on-the-spot street identity checks outside that basis.

Oversight

CNIL (French data protection authority); internal Gendarmerie monitoring tool 'Citar' introduced December 2025; no dedicated facial recognition legislation exists

What we know

French police and gendarmerie officers carry handheld devices, branded NEO (nouvel equipement operationnel), that run facial recognition software developed by the German company Cognitec, matching photographs taken in the field against the TAJ criminal records database. Facial matching became available on these devices from 2025; a desktop version had existed since 2022. TAJ (Traitement des Antecedents Judiciaires) holds around 17 million records on arrested and accused individuals, 48 million on victims, and roughly 9 million frontal photographs, operated by the Ministry of the Interior. Queries against it rose from 375,000 in 2019 to nearly 1 million by 2024. An investigation by the non-profit outlet Disclose, published in partnership with La Quadrature du Net in March 2026, found the database being used for real-time identity checks on public streets, which the French Code of Criminal Procedure does not permit: TAJ may only be used by specially authorised officers during an investigation, not for on-the-spot verification. Disclose spoke to six individuals in Paris, Marseille and Lyon photographed by police in the preceding four years, in some cases without consent, including one person scanned at a pro-Palestinian protest in Paris. A 2023 report from the police's own internal inspectorate, the IGPN, had already found the database was 'very frequently used on public streets during identity checks'. The Gendarmerie disputes the characterisation. An editorial in its own newspaper, L'Essor de la Gendarmerie, states the tool is limited to criminal investigations, identity checks where a person lacks or refuses identification, and intelligence gathering, describing its purpose as an investigative aid rather than mass surveillance. An internal memo states that results 'must be carefully reviewed by users' and 'must not be considered as evidence'. The Gendarmerie states access is restricted to judicially authorised officers, excluding volunteer deputy gendarmes and reservists. TAJ is monitored by the French data protection authority CNIL, and the Gendarmerie introduced an internal monitoring tool called Citar in December 2025 intended to reduce unauthorised access. La Quadrature du Net, which has previously and unsuccessfully sued over French police use of facial recognition, argues there is no legislation specifying how the technology may be used, by whom, or under what oversight, and that equipping thousands of officers with it amounts to surveillance authorised by the state without a legal framework. This site records a genuine dispute between an internal memo describing careful, restricted use and independent reporting describing frequent use outside the permitted legal basis, rather than resolving it.

Source

Disclose; La Quadrature du Net; Biometric Update; L'Essor de la Gendarmerie · last checked 2026-08-30

Deployment

What we know

The Turkish national police operate PTS, a registration plate reading system, and have built average-speed violation detection on top of it. The force's own announcement of 28 February 2025 states that 80 average-speed detection systems have been installed across 15 provinces — Aksaray, Ankara, Balikesir, Bursa, Corum, Eskisehir, Karaman, Kayseri, Kirikkale, Kirsehir, Konya, Nevsehir, Nigde, Sivas and Yozgat — covering 4,035 kilometres of road, with enforcement beginning on 1 March 2025 after warning signage was installed. The announcement gives a planned national expansion to 468 systems across 54 provinces, and situates the programme in the 2021-2030 Road Traffic Safety Strategy and a Vision Zero approach. Scope note. The source read establishes plate recognition operated by the national police for traffic enforcement. It does not establish that PTS reads are used for crime detection, watchlist matching or vehicle-of-interest alerting, and no such use should be inferred from this record. Turkish reporting and vendor material describe a wider City Security Management System (KGYS) integrating plate and face recognition, but that rests on supplier statements and is not recorded here. Fields left empty and why. 'since' is omitted: the announcement dates the start of average-speed enforcement to 1 March 2025 but does not date PTS itself, which is older. 'vendor' is omitted: the announcement names no supplier. 'legalBasis' and 'oversight' are omitted: neither is established by the page read, which cites a strategy document rather than a statute. Evidence base. A publication of the Emniyet Genel Mudurlugu, read twice through differently encoded URLs with consistent figures on both passes. Turkish-language source, translated here.

Source

Emniyet Genel Mudurlugu (Turkish National Police) · last checked 2026-09-05

Deployment

Legal basis & oversight

Legal basis

Not specified in source; subject to an ICO compliance audit

Oversight

Information Commissioner's Office (consensual audit, executive summary published 18 August 2026); reasonable assurance for live use, LIMITED assurance for retrospective use

What we know

Greater Manchester Police uses both live facial recognition and retrospective facial recognition. The Information Commissioner's Office audited the force between June 2025 and March 2026 as one of five in a consensual audit programme, publishing its executive summary on 18 August 2026. The two ratings differ, and the gap is the substantive finding. GMP received reasonable assurance of compliance for live facial recognition, but only LIMITED assurance for retrospective facial recognition. The ICO attributed that lower rating to the absence of a policy governing the force's use of the Police National Database, and to a lack of defined roles for using the technology. That is the clearest single illustration of the audit programme's broader finding that compliance is weaker for retrospective searching than for live deployment, despite live deployment attracting far more public and legal scrutiny. On accuracy, the force states that in more than one million scans a single individual has been misidentified, a Caucasian woman at the Manchester Marathon. That figure is the force's own. GMP says officers are present at deployments to explain the technology to the public, that alerts are reviewed by trained officers before any action is taken, and that data from non-matches is deleted within seconds. Across all five audited forces the ICO made 107 recommendations, all accepted or partially accepted, and each force has submitted an action plan.

Source

ICO audit executive summary; Computer Weekly; Biometric Update · last checked 2026-08-28

Deployment

Legal basis & oversight

Legal basis

Not specified in source; subject to an ICO compliance audit

Oversight

ICO (Information Commissioner's Office) -- audited as part of a five-force programme

What we know

Confirmed use of facial recognition technology via ICO proactive audit (executive summary published 2025-2026), one of five forces audited alongside Essex Police, Leicestershire Police, West Yorkshire Police and Greater Manchester Police. The ICO's blog covers both live and retrospective facial recognition without breaking down which variant each force uses -- recorded here as unspecified pending the individual audit summary, to avoid the classification error previously corrected for the Toronto/York-Peel entries. 107 recommendations were made across the five forces; all accepted or partially accepted. Four areas flagged as needing urgent attention: senior oversight/accountability/training; record-keeping on what data is used and shared; ensuring retrospective FR images come from appropriate sources and aren't over-retained; and checking systems for accuracy/bias. A final audit of the Metropolitan Police is scheduled for later in 2026.

Source

ICO, 'Facial recognition in policing: earning public trust through strong data protection governance', 18 August 2026 · last checked 2026-08-24

Deployment

Legal basis & oversight

Legal basis

State police laws. On 16 February 2023 the Federal Constitutional Court held the enabling provisions in Hessian and Hamburg police law unconstitutional in their then form, ruling that automated analysis of stored data is itself an interference with informational self-determination requiring a proportionate legal threshold.

Oversight

Federal Constitutional Court; state legislatures and state data protection authorities

What we know

Hesse has operated hessenDATA since 2017, with automated access to police information, case management and preventive data systems, telecoms and radio-cell data, mobile phone forensic extractions and police telex traffic, plus social network data subject to prior court order. Bavaria's VeRA went live on 25 December 2024 after pilot operation, with roughly 100 uses reported by May 2025. North Rhine-Westphalia operates DAR and Baden-Württemberg had planned adoption. The Constitutional Court described the capability as able to generate new suspicions and information police had no prior grounds to look for, and potentially to build movement, behavioural, relationship and personality profiles or machine-generated dangerousness assessments in the manner of predictive policing. The Bundesrat has called for a jointly financed federal interim platform. Critics note that while source code sits on German servers there is no assurance copies cannot be moved to the United States, and that Palantir was reportedly given unrestricted access to Bavarian police files for integration.

Source

Bundesverfassungsgericht judgment 1 BvR 1547/19; Deutsche Welle; heise · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Section 14 of the Hessian Security and Order Act (HSOG), as amended in December 2024 and in force from 2 February 2025. Real-time biometric identification sits at Sec. 14(8) and 14(9); judicial authorisation is required under Sec. 14(11); logging of each use is required under Sec. 14(10).

Oversight

Judicial authorisation required in principle before each use. Where the police authority orders use itself, judicial confirmation must be obtained within 24 hours. Each deployment must be logged under Sec. 14(10) HSOG.

What we know

Hessen Police states its system compares the biometric facial characteristics of a specifically wanted person against people appearing in designated video-protection zones, to locate that person, a missing person, or a victim in defined circumstances. Non-matches are stated to be deleted immediately, and matches are reviewed by police personnel rather than acted on automatically. This is Germany's first entry on this tracker. Correction and detail, verified August 2026. The start date is 10 July 2025, not 2026: the Hessian Interior Ministry confirmed Frankfurt police began biometric real-time remote identification in the Bahnhofsviertel on that date, making Hesse the first German federal state to deploy live facial recognition. Reporting cited by hessenschau put the deployment at around 50 cameras in that district. The statutory scope is narrower than a standing watchlist model. Section 14(9) HSOG permits targeted search for a specific person only where necessary to avert an actual and present or actual and foreseeable danger, a terrorist offence, or to search for victims of kidnapping, human trafficking or sexual exploitation, or for missing persons recorded in police information and search systems. Section 14(8) additionally allows an individual already flagged by pattern-recognition alerting to be identified where there is danger to life or limb. Use must be limited in time and place to what is strictly necessary, judicial authorisation is required in principle, and each use must be logged. How the legal basis arose is itself contested and worth recording: the facial recognition and pattern-recognition powers were added to the Bill by a short-notice amendment on 5 December 2024, after the public expert hearing on the legislation had already taken place on 12 November 2024. Critics including netzpolitik.org note that Hessian video surveillance is not confined to statistically crime-affected locations but may also cover police-defined 'Angstraeume' (spaces perceived as unsafe), and raise the standard accuracy-disparity objection that facial recognition produces more errors for women and people of colour.

Source

Polizei Hessen, own published guidance; Hessian Ministry of the Interior · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Ministry of Justice and HMPPS policy; no dedicated statute

Oversight

Ministry of Justice has published revalidation studies of the actuarial instruments

What we know

Combines a manual assessment covering offence analysis, accommodation, education and employment, finances, relationships, lifestyle, drugs, alcohol, emotional wellbeing, thinking and behaviour, attitudes and health, with statistical calculators. OGRS 3 uses gender, previous sanctions, age at current conviction, age at first sanction and offence type to output a percentage likelihood of proven reoffending at one and two years; OGP and OVP combine that score with questionnaire sub-scores for general and violent reoffending predictions. Used on more than 1,300 people per day, and in use since 2001 with rollout across the prison and probation system between 2001 and 2005. Outputs feed sentence planning, prison categorisation and supervision decisions. A replacement, Assess Risks Needs and Strengths, is being built in-house by Ministry of Justice Digital. Concerns about accuracy and discrimination have been raised by Statewatch and in trade reporting.

Source

Statewatch; Ministry of Justice revalidation studies · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

The Crime and Policing Act 2026 (Royal Assent 29 April 2026), section 181, creates an authorised person gateway for police access to DVLA driver records including the photograph, via the Law Enforcement Data Service, expanding access beyond road traffic matters.

Oversight

Biometrics Commissioner; ICO, which reported a bias finding and is monitoring a planned algorithm replacement; subject of long-running custody image retention criticism

What we know

The Home Office states the Act does not itself create a process for searching the DVLA image gallery to identify unknown individuals, and that which non-police galleries may be searched is deferred to a forthcoming Police Reform Bill. Over 25,000 retrospective facial recognition searches are run each month against a pool of more than 19 million custody images. The statutory position changed in 2026 and any account citing only the absence of a framework is now stale. The Crime and Policing Act 2026 opened police access to DVLA driver records including photographs; the Lords rejected an amendment to block facial recognition searching of the roughly 50 million image DVLA database by 123 votes to 40 in March 2026. The substantive framework for which galleries may be searched has been deferred to a further bill. The National Physical Laboratory published an accuracy and equitability evaluation of Cognitec FaceVACS-DBScan ID v5.5 in December 2025, and the ICO's 18 August 2026 statement is explicit about the finding: the algorithm's bias increases the likelihood of incorrect matches for people in some demographic groups. The Home Office and NPCC have put mitigations in place (staff training, oversight reporting, equality impact assessments) alongside plans to replace the algorithm. The ICO continues to engage with HMICFRS, the Home Office and NPCC on the replacement and is reserving the right to take further regulatory action if warranted. A Home Office consultation on the facial recognition framework closed on 12 February 2026 with a government response promised within 12 weeks. That response remained unpublished more than three months past its due date as of August 2026; a July 2026 Lords session shows the Home Secretary being directly pressed on the delay.

Source

Biometrics Commissioner annual reports · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No dedicated statute identified in the sources read. The HKPF states the scheme operates under the auspices of the Task Force on District Governance, that cameras capture public places only, and that it complies with the relevant laws of Hong Kong.

Oversight

HKPF states it maintains close liaison with the Office of the Privacy Commissioner for Personal Data and commissioned independent consultancies for Privacy Impact and Security Risk Assessments. The February 2026 LegCo paper budgets for a Privacy Impact Assessment but sets out no independent oversight body for the analytics.

What we know

Established by the Legislative Council Panel on Security paper of 24 February 2026, read in this run: 'The SmartView system is currently equipped with artificial intelligence (AI)-powered functions, such as crowd size analysis and automatic number plate recognition.' That present-tense statement in a paper laid before the legislature is why this record is operational rather than proposed. Scale, from the same paper: about 5,000 CCTV cameras installed by the HKPF and about 6,000 cameras of other government departments and organisations connected; the programme has assisted the HKPF in detecting 899 cases, including serious crimes such as murder and robbery. Over 20,000 cameras are to be added annually to reach about 60,000 across the city in 2028, with about 6,500 more by 2031. Non-recurrent cost is put at HK$4.066 billion across 2026-27 to 2031-32. The paper also states the HKPF plans to build an AI-powered video analytics toolbox so that 'the Police will be able to track and search for target persons and vehicles in a speedier and more precise manner'. That toolbox is a plan, not a current capability, and is not what this record asserts. Facial recognition is deliberately NOT recorded here. It appears nowhere in the LegCo paper and nowhere on the HKPF SmartView page, both of which were read in this run. Senior officers have been reported as intending to add it during 2026, but an intention reported in the press is not a deployment. See the accompanying report. Fields left out: 'since' is omitted because no source read in this run establishes when the AI functions specifically began, as distinct from when the camera programme launched. 'vendor' is omitted because the paper describes a direct engagement procurement approach without naming the supplier. 'techSlug' is omitted because no existing catalogue entry fits; a SmartView technology record is proposed in the same changeset, and the slug can be linked in the admin panel after it is applied.

Source

Legislative Council Panel on Security · last checked 2026-09-07

Deployment

Legal basis & oversight

Legal basis

No dedicated statute in most jurisdictions; typically procured under standard municipal contracting

Oversight

Varies by city; several cities have run independent audits of accuracy and cost-effectiveness

What we know

Acoustic sensors distributed across a coverage area analyse sound signatures to detect and locate probable gunfire in real time, alerting police. One of the most widely deployed AI-enabled policing technologies in the US, though this entry was missing from the tracker despite the site's existing gunshot-detection topic tag anticipating it. Accuracy and cost-effectiveness have both been contested in independent city audits in several jurisdictions.

Source

SoundThinking company disclosures and multiple city audits · last checked 2026-08-18

Deployment

Legal basis & oversight

Legal basis

2025 amendments to Hungarian law, in force 15 April 2025, extending permitted FR use from custodial-sentence offences to all infraction proceedings

Oversight

None domestically; subject of complaints to the European Commission

What we know

Three amendments passed by Hungary's Parliament within 24 hours in March 2025 without public debate extended facial recognition from serious offences to ALL infraction procedures -- the examples given by challengers being identifying Pride march attendees or jaywalkers. The same legislative package created a legal basis to ban Pride marches. Cameras were installed along the Budapest Pride route in June 2025 (reportedly the largest in Hungarian history); police announced 7 July 2025 that no participants would be fined. Civil society coalition (Civil Liberties Union for Europe, EDRi, ECNL, HCLU) argues this meets the EU AI Act's definition of a prohibited 'real-time' remote biometric identification system rather than merely high-risk 'post' identification, and wrote to the European Commission urging infringement proceedings; 17 EU member states have formally accused Hungary of contravening EU values over the wider package. A peer-reviewed article appears in the Maastricht Journal of European and Comparative Law (July 2026). Whether the Commission has opened AI Act infringement proceedings specifically was not established in this research pass.

Source

EDRi; European Center for Not-for-Profit Law; Maastricht Journal of European and Comparative Law · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No dedicated statute; deployed without a completed Privacy Impact Assessment

Oversight

DHS OIG audit opened Jan 2026 after Senators Warner and Kaine raised concerns; no completed PIA

What we know

Smartphone app letting ICE/CBP officers photograph a person during a street encounter and match against CBP's Traveler Verification Service (same system used for airport processing), with contactless fingerprint capture against IDENT as a fallback if no facial match returns. A 'Super Query' feature reportedly searches databases covering individuals, vehicles, aircraft, vessels and addresses simultaneously. Used over 100,000 times since launch per a January 2026 Illinois/Chicago lawsuit. A Privacy Threshold Analysis states ICE does not offer the option to decline biometric collection, retains photos and fingerprints for 15 years including non-matches, and may collect regardless of citizenship or immigration status. May be extended to 1,000+ state and local 287(g) agencies. Differs from UK operator-initiated facial recognition in three ways: it queries border/travel biometric infrastructure rather than a policing watchlist (so includes people never arrested), adds fingerprint capture as a fallback, and operates federally -- bypassing the 20+ US city and state facial recognition bans. On 5 Feb 2026 the 'ICE Out of My Face Act' was introduced to ban ICE/CBP biometric use; a coalition including EFF and EPIC has demanded DHS shut the app down.

Source

ACLU; Senator Markey follow-up letter to ICE; Immigration Policy Tracking Project · last checked 2026-08-24

Deployment

Legal basis & oversight

Legal basis

Police Ordinance Amendment No. 40 (Special Photography Systems), passed by the Knesset on 14 January 2024, authorises police camera systems that identify an object such as a vehicle registration plate. The amendment expressly excludes biometric identification.

Oversight

Implementing regulations approved by the Knesset National Security Committee in September 2024, with a two-year retention period for images of occupants and passers-by. Three Supreme Court petitions by the Association for Civil Rights in Israel; on 14 July 2026 the Court issued a conditional order requiring the State to justify contested provisions, before an expanded seven-justice panel.

What we know

A national system that photographs vehicles on Israeli roads, decodes the registration plate and photographs the vehicle's occupants. The Association for Civil Rights in Israel describes it in Hebrew as an automated system that tracks and records the movements of citizens travelling on the country's roads. It was in use before any statutory authority existed: ACRI's first Supreme Court petition was filed in January 2021 and was dismissed by consent in March 2024 once the Knesset had legislated. A second petition, filed August 2024, was dismissed in November 2024 with the petitioners directed to exhaust the regulations process. A third, filed March 2025, is live, and on 14 July 2026 the Court issued a conditional order against the State before an expanded panel of seven justices. What the law does and does not cover matters here. Amendment No. 40 authorises identification of an 'object' — a registration plate — for a closed list of purposes: preventing or detecting serious offences and offences endangering public safety or state security, preventing severe harm to person or property in the face of real danger, locating missing persons, and enforcing exclusion and restraining orders. It does not authorise biometric identification, and a facial recognition clause was removed during passage. Fields left empty and why. 'since' is omitted: no source read establishes when the system began operating, only that it predates the January 2021 petition. 'vendor' is omitted: no supplier is named in any source read. Evidence base. This entry rests on the Association for Civil Rights in Israel, a responsible litigant organisation reporting proceedings it is party to, read in both English and Hebrew; on law.co.il's report of the Knesset's second and third readings on 14 January 2024; and on Shakuf's report of 19 September 2024 on the approval of the implementing regulations and the two-year retention rule. The Hebrew sources were read in the original and are translated here. The text of Amendment No. 40 itself was not read directly, and the legal basis field should be treated as reported rather than quoted.

Source

Association for Civil Rights in Israel · last checked 2026-09-05

Deployment

Legal basis & oversight

Legal basis

Legislative Decree 51/2018 (Italy's implementation of the EU Law Enforcement Directive). The Garante treated SARI Enterprise as automating an existing AFIS search rather than creating a new category of processing, so it rests on the same legal basis as the pre-existing manual search.

Oversight

Garante per la protezione dei dati personali issued a favourable opinion in 2018. No dedicated ongoing independent oversight of the AFIS-SSA database composition or of SARI Enterprise use has been identified.

What we know

SARI Enterprise performs retrospective facial recognition searches, submitting a facial image against the AFIS photographic database and returning a ranked candidate list for human examination. Approved by Italy's data protection authority, the Garante, in 2018, and described in Italian police documentation as an established system used by the Scientific Police. A point the police's own explanation makes explicitly, and worth recording because it is the distinction this tracker exists to draw: the result is not itself a legally determinative identification. A trained operator assesses the candidates, and the output is an investigative lead rather than courtroom evidence. Recorded separately from SARI Real Time, which is a distinct live version of the system with a different and unresolved legal status. Sourcing and detail, verified August 2026. The Garante's approval is provvedimento n. 440 of 26 July 2018 (doc. web 9040256). Its reasoning is worth recording precisely: the Garante concluded SARI Enterprise did not constitute a new processing operation distinct from the manual AFIS search it replaced, but rather automated operations that previously required manual entry of identifying features, returning a priority-ordered list of similar photographic records. That reasoning is why the system rests on the same legal basis as the manual process rather than requiring its own. Scale of the reference database: in a written parliamentary response the Interior Ministry stated AFIS then held 17,592,769 photographic records corresponding to 9,882,490 individuals. The database includes people charged with offences in Italy, but also irregular migrants and anyone required to renew or convert a residence permit, which means the searchable gallery extends well beyond people suspected of crime. Contested, and recorded because it bears directly on verifiability: a FOIA-based investigation by StraLi and Privacy Network running from April 2025 to April 2026 reported that the Interior Ministry does not measure the system's effectiveness or its error rate, and criticised the 2018 Garante opinion as brief, not setting out its reasoning in full, and citing documents that were never made public. Italy has separately extended a moratorium on facial recognition in public places pending organic legislation that has still not been passed.

Source

Garante per la protezione dei dati personali, provvedimento n. 440 of 26 July 2018 · last checked 2026-08-27

Deployment

What we know

A place-based predictive policing system that identifies clusters of recorded burglaries and flags areas where further offences are considered likely, used to direct patrols. Established by three independent sources read in this run and by no cantonal police publication, which is why the confidence rests on the multiple-sources limb rather than on official evidence. SRF, the Swiss public broadcaster, reported on 27 April 2020 that the force had used the software for five years and quoted its media spokesman Roland Pfister saying 'Die Software fasst keine Einbrecher' - the software does not catch burglars - and describing its value as putting officers in the right place, primarily for prevention. AlgorithmWatch CH's atlas of algorithms in Swiss public administration records Precobs in use in Switzerland since 2013 across Zurich, Aargau and Basel-Landschaft, names LogObject AG as vendor, and records Basel-Landschaft discontinuing it in 2020 on the ground that it no longer met the increased requirements of crime analysis. Netzwoche, 1 July 2026, reports the Aargau force still using Precobs alongside a second tool called Picar, and states that these process case data rather than personal data. The 'since' field is left empty because those sources do not agree on a start date for Aargau specifically: SRF's 2020 account implies roughly 2015, while the AlgorithmWatch entry dates Precobs in Switzerland as a whole to 2013. Legal basis and oversight left empty: no source read establishes either for this force. techSlug omitted deliberately - see the accompanying report.

Source

SRF (Swiss Radio and Television) · last checked 2026-09-07

Deployment

What we know

Established by multiple independent reports of the same police announcement, four of which were read in this run: The Korea Herald and the Korean-language AI Times, both from 5 March 2024, and Financial News and Newspim, both from 13 August 2026. What the March 2024 sources establish: the National Office of Investigation put a deepfake detection tool into use, analysing a video suspected of face-swapping and returning a judgement on authenticity in five to ten minutes; a detection rate of about 80 per cent; and a training corpus of 5.2 million data points drawn from 5,400 people, including one million Korean and 130,000 other Asian samples. Officials stated at the time that because the detection rate is not complete, results guide the direction of an investigation rather than serve as evidence. What the August 2026 sources add, and why this is recorded as operational rather than announced: the software has been applied in 1,636 investigations, including 72 election crime cases in 2026, and sits within a three-part response alongside specialist forensic video analysis and AI forensics that traces which generative program produced a file. Police reported 419 people arrested for deepfake offences between January and July 2026, with 17 detained. Those arrest figures are the output of the investigations, not of the software alone, and should not be read as a measure of its accuracy. 'vendor' is deliberately omitted. The Korean sources describe the model as domestically developed and credit the National Office of Investigation, naming no commercial supplier, while a 2024 vendor press release separately announced a deepfake detection solution developed in collaboration with the agency. This run could not establish whether these are the same system, and averaging the two would be a guess. All four sources read for this record are Korean or Korean-English media reporting an official announcement; no primary police release was located. 'legalBasis' and 'oversight' are omitted because no source establishes either. 'techSlug' is omitted because no catalogue entry covers synthetic media detection; a technology record is proposed in the same changeset.

Source

The Korea Herald · last checked 2026-09-07

Deployment

Legal basis & oversight

Legal basis

Not specified in source; subject to an ICO compliance audit

Oversight

ICO (Information Commissioner's Office) -- audited as part of a five-force programme

What we know

Confirmed use of facial recognition technology via ICO proactive audit (executive summary published 2025-2026), one of five forces audited alongside Gwent Police, Essex Police, West Yorkshire Police and Greater Manchester Police. The ICO's blog covers both live and retrospective facial recognition without breaking down which variant each force uses -- recorded here as unspecified pending the individual audit summary, to avoid the classification error previously corrected for the Toronto/York-Peel entries. 107 recommendations were made across the five forces; all accepted or partially accepted. Four areas flagged as needing urgent attention: senior oversight/accountability/training; record-keeping on what data is used and shared; ensuring retrospective FR images come from appropriate sources and aren't over-retained; and checking systems for accuracy/bias. A final audit of the Metropolitan Police is scheduled for later in 2026.

Source

ICO, 'Facial recognition in policing: earning public trust through strong data protection governance', 18 August 2026 · last checked 2026-08-24

Deployment

Legal basis & oversight

Legal basis

None stated

Oversight

None identified; the ACLU of Massachusetts previously demanded disclosure of the force's Spot testing

What we know

The Bomb Squad operates two Spot platforms alongside its Special Tactical Operations Team, sending them to barricaded-suspect and potential hostage incidents, to investigate suspicious packages and to help secure crowded events, feeding situational awareness back to troopers and the incident commander so people do not have to approach on foot. Deployments have included a barricaded-suspect shooting scene in Danvers, Boston's Independence Day celebration and the Boston Marathon. Spot is a teleoperated mobility platform carrying cameras, roughly 75 pounds and costing around 100,000 dollars, not an autonomous decision-making system. The NYPD separately suspended Spot in 2021 after public backlash, then acquired two units in 2023.

Source

Bloomberg; Boston Globe; Police1 · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Force policy under the Data Protection Act 2018, UK GDPR and the Human Rights Act 1998, following the Surveillance Camera Code of Practice. No dedicated statutory basis.

What we know

The force states plainly on its own transparency page that it "uses Live Facial Recognition Technology (LFR) to prevent and detect crime and help protect the vulnerable", and publishes scheduled deployments in advance — Southport town centre on 27 August 2026 and Birkenhead town centre on 5 September 2026. The start date is not established. The force page carries forthcoming and recent deployments rather than a full historical register, and the earliest date visible on it is 27 August 2026. Secondary reporting places a first Liverpool city centre deployment in December 2025, but that was not confirmed against a force publication and the field is deliberately left empty rather than filled from an unverified figure. No equipment supplier is named on the force page. The page refers to National Physical Laboratory testing of the algorithm without identifying the manufacturer.

Source

Merseyside Police live facial recognition transparency page · last checked 2026-09-04

Deployment

Legal basis & oversight

Legal basis

No dedicated statute. Governing authority is now R (Thompson and Carlo) v Commissioner of Police of the Metropolis (High Court, 21 April 2026), read with R (Bridges) v South Wales Police (Court of Appeal, 2020).

Oversight

Force LFR policy documents; ICO engagement

What we know

In Thompson the High Court dismissed a judicial review brought by Shaun Thompson and Big Brother Watch's Silkie Carlo, upholding the Met's 2024 LFR policy as lawful and compatible with Article 8 and data protection law. Any account of UK live facial recognition citing only Bridges is now stale. Deployed at targeted operations rather than continuously, with published deployment records and arrest figures. The force says it has arrested more than 2,000 people using LFR since the start of 2024. The deployment model is changing materially. The Croydon fixed-camera pilot, running October 2025 to May 2026, produced 173 arrests from more than 470,000 passers-by with one false alert on the Met's own figures. In June 2026 the Met confirmed fixed LFR cameras would be installed in the West End and Soho before the end of 2026, with six further areas during 2027, and in August 2026 the Mayor of London indicated the Oxford Street cameras would become permanent. This is a shift from mobile, targeted, time-limited deployments towards standing infrastructure, and is arguably a more significant development than any single deployment figure. Use is spreading nationally: Statewatch's June 2026 survey recorded 12 forces using LFR, with trade reporting citing 13 of 43; treat the figure as at least 12 to 13 forces as of mid-2026 and rising. The Home Office announced 40 additional LFR vans in January 2026. First protest use came in May 2026. An investigation by Liberty Investigates and The Times found children had featured at least 1,600 times on live facial recognition watchlists across six forces, including children as young as 12; the UK Joint Committee on Human Rights has formally questioned the Home Secretary about it. Research by London Assembly Member Zoe Garbett found over half of London deployments took place in areas with a higher proportion of Black residents than the city average. Static deployment, September 2026. On 3 September 2026 the Westminster Central West BCU announced on its own X account that it was deploying live facial recognition to crime hotspots in Westminster borough that day. Commissioner Sir Mark Rowley has said cameras will be fitted across the West End and Soho by the end of 2026. These are static during deployments and attached to street furniture rather than van-mounted, though movable between deployments. It follows a six-month Croydon pilot the Met says ran October 2025 to March 2026, producing 173 arrests across 24 operations. In April 2026 the High Court dismissed a judicial review of the Met's facial recognition policy in R (Thompson and Carlo) v Commissioner of Police of the Metropolis, finding the specific policy compatible with the relevant human rights requirements. The judgment is subject to a proposed appeal. Biometrics commissioners warned in May 2026 that dedicated legislation could still be three years away, and the February 2026 Home Office consultation had produced no published outcome by September.

Source

Metropolitan Police; IBTimes UK; Liberty Investigates and The Times; London Assembly · last checked 2026-09-07

Deployment

Legal basis & oversight

Legal basis

Extraction powers derive from PACE and digital extraction provisions in later legislation; the procurement notices themselves do not set out the legal basis

Oversight

The Forensic Science Regulator would be the relevant UK body but is not named in the procurement documentation

What we know

The Mayor's Office for Policing and Crime is extending a contract for 10 Cellebrite Premium tools in Metropolitan Police digital forensics hubs and at Lambeth headquarters, valued at over 5.4 million pounds. Greater Manchester Police confirmed procurement of the Cellebrite Inseyets suite, reported at 3.4 million pounds, with plans for eight regional hubs. The Foreign Office, Home Office, Ministry of Defence and at least 25 other UK forces are reported to have Cellebrite partnerships. Deployment follows a tiered model, with level one extraction at police stations and level two in force digital hubs. A scope caveat recorded deliberately: the core product is device extraction, which is digital forensics infrastructure rather than an algorithmic decision system. The newer analysis layer adds AI capabilities but has no named force deployment. Cellebrite tools were reportedly used to access two journalists' phones by the Police Service of Northern Ireland in 2018, and a journalist's phone in Serbia.

Source

UK public procurement notices (Find a Tender); trade reporting for the GMP figures · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No facial-recognition-specific statute identified. No federal legislation mandating disclosure in criminal proceedings (the proposed Justice in Forensic Algorithms Act) has passed.

Oversight

DC Superior Court order (7 July 2026, Judge Neal Kravitz) compelling prosecutorial disclosure of the identification pipeline in one specific case; no standing regulatory oversight body identified

What we know

A Metropolitan Police Department investigation into a non-fatal shooting on 8 June 2026 near Howard University led to the arrest of Marquis Foster, 43, on 1 July, charged with assault with intent to kill while armed. Investigators identified Foster by feeding surveillance footage from the crime scene into Clearview AI, which matches a submitted image against a database the company describes as holding public photographs scraped from social media and other websites. Sources differ on the database's scale: DC Witness, reporting directly from the court, states over 30 billion images; Biometric Update's coverage of the same proceedings states more than 70 billion. Both figures are reported here rather than resolved. DC Superior Court Judge Neal Kravitz ordered prosecutors on 7 July 2026 to disclose to both the court and the defence exactly how the technology was used, after defence counsel Elizabeth Weller told the hearing she had not previously encountered the process. The parties were scheduled to reconvene on 22 July. The order does not itself rule on the admissibility or reliability of the identification; it compels disclosure of the identification pipeline, which civil liberties attorneys have described as an opaque, multi-step process that is difficult to challenge without exactly this kind of order. This is reported as part of a wider pattern: US commentators identify at least eight documented wrongful arrests attributed to facial recognition in 2026, including cases in Florida and Nevada involving different vendors and forces, though the Foster case itself does not allege a wrongful identification, only seeks transparency into how the identification was made. No federal legislation comparable to the proposed Justice in Forensic Algorithms Act, which would mandate disclosure in criminal proceedings nationally, has passed.

Source

DC Witness; Biometric Update · last checked 2026-09-03

Deployment

What we know

Established by the prosecution service's own signed procurement instrument, Disposición OAF 36/2025 of 11 June 2025, actuación interna 30-00103493, which awards Contratación Directa por Exclusividad 3/2025 to VEC S.R.L. (CUIT 30-61564727-2) for 'la Licencia de software CLEARVIEW PRO por un periodo de 12 meses para uso del Ministerio Público Fiscal de la CABA' at USD 27,588 including VAT. VEC S.R.L. is the exclusive distributor; the vendor field records Clearview AI as the maker of the software. elDiarioAR reported in 2025 that the tool was in use by August 2025 and that the prosecution service states it is used only within a criminal investigation at the express request of a fiscal, to shorten the work of tracing images already published on the public internet, and 'not for preventive purposes nor intelligence tasks'. Unlike the city's separate street camera system, this tool works on images from social media and the open web rather than live camera feeds. legalBasis is empty: the disposición is a procurement instrument, not a legal basis for biometric matching, and no source read establishes a statutory framework or a requirement of prior judicial authorisation — the safeguards described are the prosecution service's own internal statements. oversight is empty for the same reason. CELS and ODIA, quoted by elDiarioAR, argue the acquisition disregards both the courts and civil society given that the city's other facial recognition system remains judicially suspended, and note the Dutch data protection authority's EUR 30.5 million fine against Clearview in September 2024. The procurement document is in Spanish.

Source

Ministerio Público Fiscal CABA · last checked 2026-09-10

Deployment

Legal basis & oversight

Legal basis

State mandate; no independent judicial authorisation process reported

Oversight

None independent identified; internal Communist Party and state administration only

What we know

A national video surveillance infrastructure linked to the Ministry of Public Security, combining facial recognition with large-scale camera networks. Skynet dates from the mid-2000s as the urban system; Sharp Eyes is the rural and community extension, notable for integrating street-level and private or household cameras into wider monitoring platforms. A regulatory split emerged in 2025 and 2026 that is worth recording precisely because of what it does not cover. The Cyberspace Administration and Ministry of Public Security's Measures for the Security Management of Facial Recognition Technology Applications took effect on 1 June 2025, and a 2026 enforcement campaign has restricted forced commercial face scanning, with fines issued to businesses. Police systems are untouched by it. The regulation constrains private-sector use while leaving state surveillance capability intact. Le Monde reported in April 2026 on public security cameras automatically identifying ordinary citizens for minor traffic infractions in Shanghai, and noted there is no official national camera count. Figures of 600 to 700 million cameras circulating in secondary coverage are estimates and should be treated as such. This entry rests substantially on human rights researchers and investigative journalism operating under significant state restriction on information. That caveat is as important as the entry itself.

Source

IPVM technical analysis; Georgetown CSET; Brookings Institution · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Internal Ministry of Internal Affairs authorisation; no independent judicial warrant process reported for routine use

Oversight

None independent identified

What we know

Built around a Unified Data Storage and Processing Center in Moscow that aggregates CCTV feeds, with NtechLab supplying the facial recognition software. Official figures reported in 2024 put the national camera count above one million, with around 230,000 in Moscow and roughly a third connected to facial recognition. The European Union placed both NtechLab and Tevian, the two companies most directly documented as having supplied the system, under human rights sanctions in July 2023 (Council Implementing Regulation (EU) 2023/1495), for involvement in serious human rights violations including arbitrary arrests and detentions. NtechLab is owned via Rostec, Russia's state military-industrial holding company. A further contributing algorithm, Kipod, was developed by Belarusian company Synesis, separately sanctioned over its use identifying protesters in Belarus. The European Court of Human Rights ruled in July 2023, in a case brought by a solo protester identified via Moscow metro CCTV, that the system's use violated Articles 8 and 10 of the Convention. Russia left the Council of Europe in 2022 and is no longer bound by ECHR rulings, so the judgment carries no domestic legal effect but remains a significant international finding. Scale claim, August 2026: Moscow mayor Sergei Sobyanin stated that more than 30,000 federally wanted people had been detained over the past year with the help of the Sphere system, now operating in 60 Russian regions. This is an official Russian claim reported by state-linked media, not independently verified, and if accurate would make this a near-national rather than a Moscow system. A change that degrades future evidence quality and should be read alongside every figure above: OVD-Info, the principal independent monitor of this system, was designated an extremist organisation by Russian authorities in June 2026 and banned from operating in Russia. Its documented case counts, including at least 595 cases recorded in the 2021 to 2022 window, are likely to be the last comprehensive independent record for some time, and future counts should be assumed to undercount. Two documented, named uses illustrate the general pattern described above. Several people who attended the funeral of opposition leader Alexei Navalny on 1 March 2024 were detained in the following days, reportedly identified via CCTV and online footage of the memorial (EU Asylum Agency country-focus report, citing Eurasianet). Separately, Human Rights Watch has documented the system's use to locate men evading military conscription following Russia's 2022 mobilisation order. Leaked internal budget documents reviewed by Estonian outlet VSquare describe a stated purpose of "recognising... individuals for destructive and/or disloyal behavior" and allocate at least 11.2 billion rubles (over EUR 111 million) to the system's development for 2024 to 2026, corroborating the scale already recorded above rather than introducing a separate figure.

Source

OVD-Info documentation; Meduza, RFE/RL and The Bureau of Investigative Journalism reporting; EU Asylum Agency; Human Rights Watch; VSquare · last checked 2026-09-03

Deployment

Legal basis & oversight

Legal basis

UK GDPR legitimate interests; ICO assessed

Oversight

ICO issued findings on Facewatch in 2023

What we know

A private facial recognition system used by retailers rather than by police, which is what makes it worth tracking here: it is a commercial deployment whose output increasingly flows to the state. That flow became direct in July 2026, when Facewatch's Crime Management Platform began automatically alerting police to repeat-offender matches across more than 125 retailers. Until then, retail facial recognition alerts reached police only through a shopkeeper choosing to report. This is a material change to the private-to-state data pathway and the most significant development on this entry. Misidentification harm is now documented repeatedly. Sainsbury's paused use at one store on 17 August 2026 after a customer was falsely accused of shoplifting, while continuing to expand the system from 55 stores towards up to 200 by the end of 2026. Guardian reporting in May 2026 documented shoppers falsely identified struggling to clear their names, with limited routes of appeal. Regulatory position has moved beyond the ICO's 2023 findings: the ICO opened an investigation in February 2026 into a case in which a B&M shopper was wrongly accused of theft. Facewatch obtained ISO/IEC 42001 AI management certification in June 2026. The February 2026 ICO investigation rests on a single regional outlet and would benefit from corroboration against an ICO statement.

Source

ICO statement on Facewatch · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Ordinary municipal procurement; no dedicated statute

Oversight

Governed by individual memoranda of understanding and local retention laws per force

What we know

Cloud platform integrating existing camera, ALPR, drone and dispatch feeds into one live operating picture for a police real-time crime centre. Its distinguishing feature, the fususCORE device, lets private camera owners (businesses, schools, households) voluntarily connect cameras to the police platform on customisable terms; some jurisdictions run donor programmes providing free units. Axon states there is no facial recognition in the platform -- a specific, checkable claim. Documented contracts: Providence RI ($750,000, signed June 2024, to at least 2028); Chula Vista CA (Jan 2026 council report, replacing Motorola Solutions -- a second entry for a force already on this tracker for Drone as First Responder); Knoxville TN ('Connect Knoxville', residents/businesses buy fususCORE devices); Lawrence KS. Related: Axon Lightpost, a streetlight-powered ALPR camera that pairs with Fusus, making Axon a direct ALPR competitor to Flock Safety (Axon was previously a Flock investor before the two split).

Source

Chula Vista City Council staff report; Stanford Law CodeX presentation transcript · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No dedicated statute. Officers are required to review and take responsibility for the final report

Oversight

Internal department review of each generated report before submission; no external oversight body identified

What we know

US Department of Justice research documents both products as operational: body-worn camera audio is transcribed and a generative AI model produces a first draft report, which the officer must review, amend as needed and sign. Both systems work from audio only, not the camera's visual footage, according to the DOJ's own description. This differs in kind from a general-purpose assistant like Microsoft Copilot, which the site already tracks separately: these are purpose-built for report drafting from a fixed audio input, not a general chat tool applied to policing tasks.

Source

US Department of Justice research documentation · last checked 2026-08-26

Deployment

What we know

Established from the National Police Agency's own White Paper on Police, in the section on strengthening police capability through AI and advanced technologies. The page is in Japanese and was read twice in this run through differently encoded URLs; both readings gave the same sequence, and the second returned the Japanese verbatim: a demonstration trial from fiscal 2019 in which the AI was trained on reports filed by specified business operators and on the results of their existing sorting and analysis, followed by the sentence stating that full operation of the mechanism began in March 2022. The system prioritises, from a reported flow of over 400,000 suspicious transaction reports a year, those requiring staff attention. It supports human analysts rather than reaching conclusions on its own. This finding rests on a translated Japanese-language primary source. The underlying document is an official police publication, which is the strongest class of source the site recognises, but the operator should be aware that no English-language original was available. Two further NPA systems described on the same page are deliberately NOT recorded: an AI tool for detecting regulated drug advertisements on social media, tested in fiscal 2021, and a system identifying suspect vehicles from surveillance footage, tested from fiscal 2019 with automatic vehicle detection added by fiscal 2021. Both are described only as demonstration trials, with no statement that either entered operation. A fourth item, age-progression facial imaging developed by the National Research Institute of Police Science, is research rather than a deployment. Fields left out: 'vendor' is omitted because none is named. 'legalBasis' and 'oversight' are omitted because the white paper section describes neither. 'techSlug' is omitted because no catalogue entry covers AI triage of financial intelligence reporting; this gap is discussed in the report.

Source

National Police Agency White Paper on Police · last checked 2026-09-04

Deployment

Legal basis & oversight

Legal basis

No specific statute authorised its use for its first eight years. The interior ministry's own inspection concluded it was used in a judicial, retrospective context rather than real-time administrative surveillance, which affects but does not resolve its legal footing

Oversight

Inter-inspectorate report (IGA, IGPN, IGGN), commissioned by the interior ministry and published October 2024

What we know

This is more precise, and more interesting, than a flat operational status conveys, and is worth recording carefully. Investigative outlet Disclose revealed in November 2023 that police and gendarmerie had used BriefCam since 2015 without a clear legal basis, including its facial recognition capability, which is illegal in French public space. The gendarmerie suspended use three days after the report; the national police suspended use in December 2023. The official inter-inspectorate report, finalised February 2024 and published October 2024, examined approximately 600 total uses and found exactly one illegal facial-recognition use. It recommended lifting the suspension and continuing to use BriefCam in investigative services until it can be replaced by a sovereign French-developed system, Systeme V, which has no facial recognition capability and was still described as under development in parliamentary answers as recently as mid-2025. This entry is recorded as operational on the basis that the report recommended resuming use pending that replacement, but the transition to Systeme V should be treated as ongoing rather than complete.

Source

Disclose (original investigation); French Ministry of the Interior inter-inspectorate report · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Federal Aviation Administration waiver permitting flight beyond visual line of sight across much of New York City with no visual observer, up to 200 feet above ground level or within 50 feet of structures

Oversight

Federal Aviation Administration waiver conditions; no dedicated civil liberties oversight identified

What we know

A drone is launched to a call for service ahead of officers, with a remote pilot flying the whole mission and no personnel on scene. Live video streams to responding officers and evidence is captured from overhead. Onboard autonomy avoids obstacles, and traffic alerting warns the operator of low-flying crewed aircraft. The FAA waiver is the enabling factor rather than the technology itself, which is why the regulatory position is recorded here in as much detail as the capability. Chula Vista Police Department in California has received authorisation for low-altitude close-proximity emergency operations, and Redmond Police Department in Washington has launched a programme. In March 2026 Skydio announced that 12 public safety agencies had received FAA approval for one pilot to fly up to four drones beyond visual line of sight.

Source

Skydio; Federal Aviation Administration approvals · last checked 2026-08-20

Deployment

Legal basis & oversight

Oversight

The Office of the Privacy Commissioner and the Police Association were consulted on the 2023 baseline audit. Police said regular auditing of staff use of ANPR platforms would be added to its internal audit and assurance programme.

What we know

Established by New Zealand Police's own news release of 26 April 2023 announcing the findings of a baseline audit into how staff access externally provided ANPR data. The audit cross-referenced over 350,000 ANPR reviews from 2018 to 2022 across two provider platforms. Assistant Commissioner Johnson is quoted saying the audit confirmed the vast majority of Police use the tool within guidelines; five officers were referred for misconduct review. Current use is confirmed by the ANPR chapter of the Police Manual, published in April 2026 on the Police website, which describes 'Police use of NPI collected by third party operators' and 'approved ANPR use'. Police's own emergent technologies page separately states Police have 'Automatic Number Plate Recognition on some cars', so the deployment covers both in-vehicle reading and access to third-party networks. Fields left out and why. 'vendor' is omitted: the April 2023 release refers only to 'two provider platforms' and names neither. Searching surfaced Auror and SaferCities/vGRID as the commonly named providers and NZ Police's own proactive release index lists an Auror review, but no page read in this run states which platforms Police uses, so nothing is asserted. 'since' is omitted: the audit covers 2018 to 2022, which establishes use by 2018 but not a start date. 'legalBasis' is omitted: no statute or provision was established by any page read. The full ANPR Police Manual chapter and the October 2025 Technology Capabilities List are both PDFs under police.govt.nz/sites/default/files/, which is disallowed to automated fetching by that site's robots.txt. Those two documents would settle vendor, legal basis and scope.

Source

New Zealand Police · last checked 2026-09-09

Deployment

Legal basis & oversight

Oversight

Police use of Facial Recognition Technology policy, published October 2025. Independent review by Dr Nessa Lynch and Dr Andrew Chen, which made ten recommendations Police say they are responding to. Expert Panel on Emergent Technologies and the Algorithm Charter.

What we know

Established by two sources read in this run. RNZ, 14 September 2024, reporting a Police response under the Official Information Act: the Automated Biometric Identification System (ABIS 2) was used 89 times between 21 March 2022 and 3 September 2024, and Police said none of these searches involved 'live' searches - all were retrospective searches to identify images of suspects. RNZ reports that Police adopted their first facial recognition policy in August 2024, restricting live use to rare situations involving immediate risk to life. Separately, New Zealand Police's own facial recognition technology page confirms a published FRT policy, an independent expert review by Dr Nessa Lynch and Dr Andrew Chen covering current and future uses, ten recommendations, and that Police hold 'planned and unused capability'. The Police Manual chapter page dates the current FRT policy document to October 2025. Status is operational because Police's own OIA response confirms searches actually run. It is explicitly retrospective: no live or real-time facial recognition deployment is established, and Police say none has occurred. Fields left out and why. 'vendor' is omitted. RNZ names Dataworks Plus as the supplier of the ABIS 2 upgrade, described as costing about 20 million dollars five years earlier, but that rests on one article; searching also surfaced NEC in connection with New Zealand Police biometrics, which appears to relate to the separate fingerprint AFIS rather than to ABIS. Rather than risk conflating two systems, the supplier is recorded here in the notes and attributed to RNZ instead of being asserted as a field. 'since' is omitted: 21 March 2022 is the start of the period the OIA response covers, not an established start date for the system, and RNZ reporting from 2020 indicates ABIS work was already under way. 'legalBasis' is omitted: no statute was established by any page read.

Source

RNZ · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

Not stated

Oversight

Not stated

What we know

Used for periods when multiple callers report the same incident, such as a road collision. The AI assesses incoming calls and identifies those relating to an incident already known to the emergency communications centre, allowing human call-takers to prioritise calls requiring intervention. Human staff remain responsible for calls needing emergency intervention. Carbyne and the district describe the technology as having processed more than 3,500 events during a 90-day period and reduced redundant calls by more than 30 percent among the calls it triaged. These are vendor and customer reported figures rather than independently validated efficacy measures. Axon announced a definitive agreement to acquire Carbyne on 4 November 2025 for $625 million, closing in Q1 2026 per Axon's FY2025 shareholder letter. Axon separately acquired Prepared (see the AI Emergency and Non-Emergency Call Triage entry) in late 2025; the two now anchor a combined product line branded Axon 911. This concentrates a striking share of the US emergency-response AI market in one vendor -- Axon now owns Carbyne, Prepared, Fusus, Draft One and Axon body cameras. A scope note: this is an emergency communications district handling police, fire and medical calls, rather than a police force. It is included because control rooms and emergency response fall within this site's subject scope.

Source

Axon press release; Axon investor relations; GovTech · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Home Office programme; no dedicated statute

Oversight

Home Office; a public register of police AI is planned, though Statewatch notes it is expected to align with FOI exemptions and completion does not appear to be compulsory

What we know

Announced in the Police Reform White Paper of January 2026 with 115 million pounds over three years, mobilising from April 2026 and formally launched on 10 June 2026, hosted by the College of Policing with 75 million pounds over three years within a wider package of around 140 million. First-year priorities include case-file assistants, disclosure assistants, crime data integrity, rapid CCTV and digital media analysis, image identification and classification, transcription and translation, AI-assisted redaction, retail crime and tool theft matching, deepfake detection, and establishing governance for Microsoft Copilot use. It also covers a public register of police AI, a national live facial recognition capability and national data standards. Large-scale pilots involving up to ten forces are planned across 2026 and 2027, with intended national scaling from 2027. An important limit: the Home Office states that responsibility for procurement and deployment remains with chief officers. This is a coordinating and evaluating body, not a central system imposed on forces. It is recorded here because it changes how individual technologies now reach forces, not because it is itself a deployed technology.

Source

GOV.UK; National Police Chiefs' Council · last checked 2026-08-20

Deployment

What we know

Argentina's national Ministry of Security acquired a Clearview AI licence in October 2025 for USD 33,500 by direct contract, according to Chequeado's review of the ministry's procurement records and to Amnesty International's report 'Estado de vigilancia: la expansión del uso de tecnologías de vigilancia masiva por el Gobierno de Argentina', which covers December 2023 to December 2025. Amnesty records that the tool is used by the Policía Federal Argentina and that the ministry states it is used only for retrospective investigation. Clearview matches a submitted photograph against a corpus described in both sources as more than 70 billion images scraped from public internet sources. The same procurement review records a Maltego open-source intelligence licence at USD 239,013, a Nosis commercial-data subscription at USD 17,000, and 69 drones with 11 ground stations at roughly USD 1 million, giving at least USD 1.2 million of surveillance technology bought in 2024-2025. legalBasis is deliberately empty: both sources describe decrees and ministerial resolutions that expanded digital monitoring powers without prior judicial authorisation, and Amnesty records that Federal Police social-media monitoring was formalised on that footing, but no source read in this run establishes a specific statutory basis for facial recognition itself, and an absence of one is a finding about the deployment rather than a legal basis. oversight is empty for the same reason. Both figures and dates here rest on Spanish-language reporting and on an NGO report summarised in Spanish; the underlying contract documents were not opened in this run. See also the separate record for the Buenos Aires City prosecution service, which holds its own, distinct Clearview licence.

Source

Chequeado · last checked 2026-09-10

Deployment

What we know

IAPol — Inteligencia Artificial Aplicada al Servicio de Policía — produces weekly forecasts of where, when and what type of crime is expected, expressed as heat maps over a 200 by 200 metre grid with intensity values from 0 to 255, broken down by the three daily shifts, with crime type assigned by a random forest classifier. It draws on police databases including SIEDCO (crime statistics, arrests, seizures, recovered vehicles), RNMC (police code infractions) and the 123 emergency call system, and is available to police stations nationwide, where it is used to distribute personnel and direct operations. Cerosetenta, the Universidad de los Andes investigative outlet, reports it was originally built as XCrime by High Tech Software S.A.S. under a contract signed in September 2020 worth about 5 billion Colombian pesos, was renamed IAPol in mid-2025, became operational in August 2025, and has a maintenance contract running to 30 November 2026. El Espectador, reporting from the police General Directorate, describes it as generating predictive heat maps from 15 police databases with real-time patrol routing, and names Colonel Óscar Andrés Bernhard Rojas as the officer responsible; that report gives a current budget allocation of 5.8 billion pesos and an earlier start in 2019 — the 2019, 2020 and 2025 dates are not reconciled by anything read here, which is why the 'since' field carries only the year operation is reported to have begun. The Universidad de los Andes public algorithm registry lists IAPol as in operation, processing personal data, with 2026 as the operational year. Criticism is part of the record: analysts at Quantil, quoted by Cerosetenta, conclude that the location and timing outputs are descriptive statistics over historical crime data rather than machine learning, and the police say no performance metrics exist because police action itself changes the outcome. legalBasis and oversight are empty: Cerosetenta reports that Colombia has no specific regulation of police AI, that a bill was introduced in May 2025 with uncertain prospects, and that no data is published on operations conducted using the system. Sources are in Spanish.

Source

Cerosetenta (Universidad de los Andes) · last checked 2026-09-10

Deployment

What we know

General Javier Gonzáles, director of the Peruvian National Police's technology and communications directorate, said on 19 November 2025 that of about 3,000 police surveillance cameras nationwide, 500 were already operating with artificial intelligence, and that the full 3,000 would be running with AI between January and February 2026. The facial recognition component matches faces against a register of about 18,000 people with outstanding arrest warrants, with a claimed accuracy of 91 per cent, and sends an alert to the mobile phones of the responsible police unit when a match is made. Monitoring runs from the C4 command centre, coordinating with municipal camera systems in Lince, Surco and Miraflores and with the Falcon intervention unit; the force said it would add about 100 officers and seven dedicated rooms. The first arrest attributed to the system, of Jack Bebeto Rengifo, was made at the C4 in November 2025, and América TV reported in 2026 that trials in shopping centres and busy streets had led to 20 people being located and detained before the technology was extended to two-megapixel cameras capturing the face of each passenger boarding participating Lima buses. A further 10,000 cameras were announced for Lima at a later phase; that expansion is an announcement and is not part of this record. status is 'operational' on the strength of the 500 AI cameras already running and the arrests made, not on the announced 3,000. legalBasis and oversight are empty: nothing read establishes a statute, regulation or supervisory body for police facial recognition in Peru, and the accuracy figure is the force's own claim rather than an independent measurement. Sources are in Spanish. Peru's data protection authority sanctioned a separate municipal facial recognition deployment in La Victoria in July 2022, which suggests a regulator with an interest but was not verified in this run.

Source

Agencia Peruana de Noticias Andina · last checked 2026-09-10

Deployment

Legal basis & oversight

Legal basis

Not disclosed. China has no independent judicial authorisation requirement for domestic police surveillance technology comparable to Western warrant systems

Oversight

None independent of the state disclosed in available reporting

What we know

A fleet of driverless police patrol vehicles equipped with 360-degree panoramic cameras, facial recognition and onboard AI processors, operating around the clock without officers aboard. Authorities describe the vehicles as identifying watchlisted individuals, fraudulent number plates and traffic or safety violations in real time, as part of a wider strategy of embedding AI into everyday policing and reducing reliance on human patrol officers.

Source

Reported via multiple outlets citing Chinese state media · last checked 2026-08-18

Deployment

Legal basis & oversight

Legal basis

Not disclosed

Oversight

None independent of the state disclosed

What we know

Two separate Chinese manufacturers have placed humanoid robots into public-facing traffic policing roles, and the distinction between them matters because their deployments and figures are frequently conflated in coverage. SUPCON Information builds the T2, a 98kg humanoid using cameras, radar and onboard computing to identify traffic violations: riders without helmets, e-bikes carrying passengers, vehicles stopping across marked lines, and pedestrians crossing against signals. Reuters reported in August 2026 that 15 T2 units had operated in Hangzhou since May 2026, issuing more than 170,000 warnings, with SUPCON claiming a fall of over 40% in monthly helmetless riding and stop-line violations and violation-recognition accuracy above 95%. Reuters stated it could not independently verify these figures and that Hangzhou traffic police did not respond to a request for comment. Reporting at the time put the wider total at nearly 50 robots across roughly eight cities in three provinces, with deployments expected to approach 200 by the end of 2026. AiMOGA Robotics, backed by carmaker Chery, builds a separate machine, the 'Intelligent Police Unit R001', reported by Xinhua as deployed alongside human officers in Wuhu, Anhui Province, from 10 January 2026. AiMOGA has discussed exporting police robots but its executives note this requires fitting local law-enforcement IT systems, obtaining certifications and complying with data-protection rules, given the machines' cameras and sensors. This is materially different from the autonomous patrol vehicles tracked separately: those are driverless mobile surveillance platforms, while these are physically embodied robotic agents placed into public-facing safety roles. All deployment counts and effectiveness figures here originate from the manufacturers and news-agency reporting rather than police procurement documentation, and should be read as company claims. No individual municipal public security bureau has published its own verified unit count.

Source

Reuters, August 2026 (syndicated); Xinhua and People's Daily, January 2026 · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Separately regulated by the ANPG executive order (bekendtgoerelse om politiets anvendelse af automatisk nummerpladegenkendelse). Rigspolitiet is named on its own website as data controller for the system, and chapter 4 of that order governs when ANPG data is deleted.

What we know

Established by the Danish National Police's own page on its processing of personal data, which states that Rigspolitiet is data controller for ANPG, that its purpose is to obtain images of number plates from cameras, and that it is separately regulated by a named executive order whose chapter 4 sets deletion rules. Scale comes from secondary Danish reporting rather than a police publication: Version2, 4 November 2020, reported roughly 250 fixed ANPG cameras across Denmark, mobile cameras on 50-100 patrol vehicles at two cameras per vehicle, around 10 transportable manned camera sets, and a replacement tender valued at between 150 and 600 million kroner. The 'since' field is deliberately empty because two sources read in this run disagree: the Danish public-sector AI registry offentlig-ai.dk dates first police use to 2015, while Version2 says the system has been in operation since 2016. No supplier is named, for the same reason - offentlig-ai.dk names ITS Teknik A/S, but a large replacement procurement was under way in 2020 and no source read in this run establishes who supplies the system today. Oversight left empty: no source read names a specific supervisory arrangement for ANPG beyond ordinary Danish data protection law.

Source

Danish National Police (politi.dk) · last checked 2026-09-06

Deployment

Legal basis & oversight

Legal basis

Rigspolitiet is named on its own website as data controller for POL-INTEL, which is regulated by executive order; chapter 4 of that order sets out when information held in POL-INTEL is deleted.

What we know

The Danish National Police's own page on its processing of personal data establishes that POL-INTEL exists, that Rigspolitiet is its data controller, that it is used to carry out cross-cutting information analyses in support of police work, and that it is regulated by an executive order governing deletion. Everything beyond that comes from secondary sources and is flagged as such. The Danish public-sector AI registry offentlig-ai.dk, read in this run, names Palantir Technologies Inc. as supplier, dates operation to 2017, describes a 'SmartSpot' function for identifying areas at higher risk of crime, lists the police registers the platform draws on including POLSAS case management, the vehicle register, criminal records and the national photograph register, and records that as of 2023 work was under way to add telephone data, ANPG number plate records and passport and driving licence registers. The same registry states that police do not use algorithms for profiling or prediction of crime at individual level. The 'since' field is left empty because 2017 rests on that registry alone and no police or ministry publication establishing the date was readable in this run: the Danish Ministry of Justice press release on the system is blocked to automated fetching at the robots.txt layer, and the Folketing Legal Affairs Committee answers on Palantir software returned 403. Those parliamentary answers are the obvious way to settle both the date and the supplier.

Source

Danish National Police (politi.dk) · last checked 2026-09-06

Deployment

Legal basis & oversight

Legal basis

Not established in sources reviewed

Oversight

None identified

What we know

State military police (distinct operator from São Paulo's city-run Smart Sampa) running facial recognition cameras via the Integrated Command and Control Center, scanning passers-by against outstanding arrest warrants and dispatching the nearest patrol on a match. Expanded before Aug 2026 with 1,300 police-car-mounted cameras using Dahua software scanning faces and plates; a police major stated plans to extend to body-worn cameras later in 2026 -- the clearest documented case of vehicle-mounted mobile facial recognition at fleet scale found in this research. Documented misidentifications in Brazil since 2020 include a Black woman arrested in Rio (2019, mistaken identity), a woman detained in Rio (Jan 2024) after the database showed her as a fugitive while she was already serving her sentence, and a Black man wrongfully detained 26 days in Bahia over a decade-old robbery by someone else. Oxford Institute of Technology and Justice records 506 AI-driven FR systems adopted in Brazil as of March 2026, with projected reach to ~87 million people.

Source

Christian Science Monitor; Oxford Institute of Technology and Justice · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Not established in sources reviewed; Brazil's LGPD is the general data protection law but no specific authorising instrument was found

Oversight

None identified; the city publishes a live public arrest tally outside the monitoring centre

What we know

Municipal (not police-force) live facial recognition across roughly 40,000 street, public-building and voluntarily-enrolled private cameras, alerting a monitoring centre staffed by police officers when a match against judicial fugitive databases reaches 90%+ confidence. Costs around $2m/month. Official transparency reports analysed by AFP show more than 8% of people identified and arrested as fugitives in the system's first year had to be released due to errors; most arrests occurred in outlying neighbourhoods among migrants from poorer regions. As of March 2026, roughly 3,000 fugitives caught via the system, with nearly 4,000 more caught in the act. The published 8% wrongful-arrest rate is the most specific error figure found for any live FR system's operator, more precise than the Met's 'one false alert' or BTP's 'zero true matches'.

Source

AFP via France 24; Christian Science Monitor · last checked 2026-08-24

Deployment

Legal basis & oversight

Legal basis

Not established; reporting has questioned whether the deployment was reviewed under Seattle's surveillance technology framework

Oversight

Seattle surveillance technology framework, with reporting questioning whether it was applied

What we know

Analyses emergency medical 911 calls in real time, providing prompts to dispatchers and assisting decisions about whether particular callers should be routed towards a nurse-staffed service rather than receive an immediate ambulance response. In use since December 2023. This is a more consequential application than transcription, and the distinction matters: the AI output influences emergency response prioritisation rather than merely recording what was said. Documented governance concerns centre on the absence of public disclosure and review under the city's surveillance technology framework, though these derive from investigative reporting rather than a regulator's finding. As with the New Orleans entry, this is a fire department rather than a police force, included because emergency response and control rooms fall within this site's scope. On the department's own effectiveness claim, the two figures it has given do not agree. A 2024 Corti press release quoted SFD Medical Director Michael Sayre crediting the technology with a 50% increase in 911 calls routed to the nurse line; a Fire Department spokesperson later corrected that to 32%. Neither figure has been independently verified. The department also confirmed the system had run without disclosure to callers and without public review, and self-determined that it did not qualify for review under Seattle's 2017 surveillance ordinance.

Source

Seattle Times investigation, via GovTech · last checked 2026-08-26

Deployment

What we know

Established by the Singapore Police Force's own account of its Police Workplan Seminar 2026, published on police.gov.sg and read twice in this run through differently encoded URLs, with consistent results both times: 'From May 2026, the drones began operating aerial patrols from eight drone pods located outside Singapore's aerodromes.' Payloads are given as high-definition thermal imaging cameras, emergency blinkers, searchlights and loudspeakers, with remote operators at the Police Operations Command Centre. One qualification on the technology link. The catalogue's drone-first-responder record describes systems that launch automatically on emergency calls to arrive ahead of ground units. SkyGuardian is used for routine aerial patrol as well as incident response, so it is broader than the US model that record is built around. The slug is the closest fit in the catalogue and is applied on that basis, not because the two are identical. Fields left out: 'vendor' is omitted because the force publication names no manufacturer. 'legalBasis' and 'oversight' are omitted because neither is described in the source; Singapore has no published police-specific drone authorisation framework that this run could verify.

Source

Singapore Police Force · last checked 2026-09-04

Deployment

What we know

Established by the Singapore Police Force's account of its Police Workplan Seminar 2026, read twice in this run. R-COP is described as an AI-assisted chatbot that helps members of the public lodge police reports at self-help kiosks. Both readings agree that it launched in October 2025 and had expanded to 21 more Neighbourhood Police Centres as of April 2026. One detail is deliberately not recorded. The two readings of the page differed on the initial launch sites, one rendering them as seven Police Land Division headquarters and the other simply as police headquarters. Because this run cannot settle which is correct, the initial site count is left out rather than asserted. 'techSlug' is omitted because no catalogue entry fits. This is not officer report writing, which the ai-report-writing record explicitly limits to turning body-worn camera audio into an officer's draft; and it is not call triage, since the ai-emergency-call-triage record was read in this run and states that it covers 999/911 call handling rather than public-facing reporting assistants. A technology record for public-facing AI report-lodging assistants is worth creating once there is more than one documented instance; that judgement is left to the operator and is discussed in the report. Fields left out: 'vendor', 'legalBasis' and 'oversight' are all omitted because the source establishes none of them.

Source

Singapore Police Force · last checked 2026-09-04

Deployment

What we know

Established by a written parliamentary reply from the Ministry of Home Affairs, dated 3 October 2023, answering a question from Mr Desmond Choo on how the police have incorporated artificial intelligence into enforcement work. The page was read twice in this run and gave the same sentence both times: 'Investigators use an AI programme to sieve out possible obscene materials from seized devices.' The programme is not named, no vendor is identified, and no accuracy or volume figures are given. The record asserts only what the reply asserts: that such a programme is in investigative use. 'techSlug' is omitted rather than linked to csam-detection-classification. The reply says 'obscene materials', which is broader than child sexual abuse material, and one reading of the page added a child abuse framing that the verbatim sentence does not support. Linking the two would import a claim the source does not make. 'since' is omitted because the reply establishes the programme was in use by October 2023 but not when it began. Note for future cycles: this record rests on a statement now nearly three years old. Nothing found in this run contradicts it, but it should be re-checked against a more recent parliamentary reply or force publication before its verified date is advanced.

Source

Ministry of Home Affairs, Singapore · last checked 2026-09-07

Deployment

Legal basis & oversight

Legal basis

Not detailed in the source reviewed

Oversight

South Australian Auditor-General, which is the source this entry rests on

What we know

The South Australian Auditor-General's own report states SAPOL was already using AI-type tools for facial recognition matched against CCTV and custody photographs, mobile phone detection cameras, and social media monitoring, with further plans for AI-assisted interview transcription and machine learning tools for identifying leads and hidden connections between cases. This entry bundles several related capabilities the Auditor-General reported together rather than treating each as a separate deployment, since the source itself does not give each one separately confirmed operational status.

Source

South Australian Auditor-General · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Common law; considered in R (Bridges) v South Wales Police

Oversight

Court of Appeal judgment 2020 required changes to policy

What we know

Deployed for public events including Six Nations rugby fixtures, with continued operations through 2026 including the Swansea Airshow in July 2026 and deployments in Cardiff and Pontypridd in June and July 2026. R (Bridges) v South Wales Police remains the governing appellate authority on this force's use and has not been disturbed. The force has spent roughly £3.5m on the programme, with false alert rates falling over time on its own figures. Separately, South Wales Police has trained over 500 officers to use a handheld facial recognition mobile app, making it a significant operator-initiated facial recognition deployment as well as a live one. The Liberty Investigates and Times investigation into children on facial recognition watchlists found South Wales Police had placed 632 people under 18 on its watchlists since 2022, including six children under 13. The force's own guidance states no one under 13 should be included, and the College of Policing notes the technology performs less reliably on younger children.

Source

R (Bridges) v Chief Constable of South Wales Police [2020] EWCA Civ 1058 · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Organic Law 1/2004 on Comprehensive Protection Measures against Gender Violence; operated by the Secretariat of State for Security, Interior Ministry

Oversight

No dedicated external regulator identified; Eticas Foundation conducted an independent reverse-engineered audit in 2021 after offering a pro-bono audit in 2018 and receiving no response

What we know

Launched 2007, updated to VioGen 5.0 in 2019, with an AI pilot to refine the model from 2022. Runs nationwide excluding the Basque Country and Catalonia, which operate separately. Officers complete two questionnaires (Police Risk Assessment/VPR and Police Risk Evolution Assessment/VPER) covering 35 indicators, producing a score sorting cases into five bands (unappreciated/negligible, low, medium, high, extreme) that determines the level of police protection assigned to a victim. Eticas describes the method as classical statistical weighting rather than machine learning. Eticas's 2021 audit found officers keep the automatic outcome in 95% of cases despite it being designed as an overridable recommendation. New York Times reporting found at least 247 women killed by partners since 2007 after VioGen assessment, at least 55 of them scored negligible or low risk. As of 2025: roughly 92,000 active cases, 83% classified low or negligible risk. This is decision-support and risk assessment rather than the narrower predictive-policing sense of forecasting where crime will occur.

Source

European Commission Interoperable Europe Portal; AlgorithmWatch; Eticas Foundation · last checked 2026-08-26

Deployment

What we know

SITIA — Sistema Integrado de Teleprotección con Inteligencia Artificial — is a programme of the Unidad de Inteligencia Artificial e Interoperabilidad de Datos in the División de Estudios y Tecnología of the Subsecretaría de Prevención del Delito, under Chile's Ministerio de Seguridad Pública. Its own site states 'Presencia en 36 comunas de la Región Metropolitana', '+ 2.000 Cámaras unificadas' and '+ 550 Lectores de patente integrados'. The plate component performs automatic detection and recognition of number plates and is integrated with Carabineros' unified vehicle warrant database, so that a vehicle with an outstanding theft report is detected automatically. President Gabriel Boric presented the system's progress on 8 July 2024 alongside the Interior Minister and both subsecretaries, saying the aim was to bring public and private camera feeds into a single platform. On 22 December 2025 the Subsecretaría launched the first municipal deployment of SITIA Patentes in Lo Barnechea, with 115 integrated cameras, and reported four vehicles with active theft warrants seized in four checkpoints during the first week; Carabineros General Jaime Velasco described the interconnection of municipalities, private entities and Carabineros as strategically important. Earlier official reporting states SITIA Patentes had already enabled Carabineros to break up ten gangs involved in violent vehicle theft in the Metropolitan Region. legalBasis and oversight are empty: no source read establishes a statute or regulation authorising the system, and the only governance body found is an advisory council of universities and experts mentioned in the July 2024 presentation, which is not oversight in the sense this field is used. Sources are in Spanish. Note that nothing read states or denies that SITIA includes facial recognition; the site describes plate recognition, camera unification and evidence management only.

Source

SITIA, Subsecretaría de Prevención del Delito · last checked 2026-09-10

Deployment

What we know

The second component of SITIA brings security cameras into one platform for real-time monitoring 'regardless of their origin or manufacturer', including feeds owned by municipalities and by private operators, and is paired with a digital, encrypted evidence platform through which prosecutors request audiovisual material. The Ministerio del Interior y Seguridad Pública announced on 7 May 2024 that 70 cameras were connected at that point with capacity in that phase for more than 1,541, drawing on the communes of Renca, Lo Prado, Santiago, Independencia, Quinta Normal, Vitacura, Estación Central and Providencia, together with the traffic control unit UOCT and private participants Autopista Vespucio Norte, Mall Plaza Norte and Arauco Estación Central. Carabineros concentrates the project, its Metropolitan zone chief saying it would maximise prevention and institutional reaction. The SITIA site now states more than 2,000 unified cameras across 36 communes of the Metropolitan Region. This record is kept separate from the SITIA Patentes plate-reading record because they are different technologies within one programme. legalBasis and oversight are empty: nothing read establishes a statutory basis or a named oversight body, and the advisory council of universities and experts mentioned in July 2024 is advisory rather than supervisory. Sources are in Spanish. Camera counts here are the government's own published figures and were not independently verified.

Source

Ministerio del Interior y Seguridad Pública · last checked 2026-09-10

Deployment

Legal basis & oversight

Legal basis

Police and Criminal Evidence Act 1984, Human Rights Act 1998, Equality Act 2010, UK GDPR and the Data Protection Act 2018, as cited by the forces. No dedicated statutory basis.

Oversight

Deployments authorised at superintendent rank or above. Both forces publish a deployment register, and Sussex additionally publishes its standard operating procedure. Deployments are notified at least seven days in advance.

What we know

A single joint programme rather than two separate deployments: the forces describe "a joint bid to obtain two live facial recognition (LFR) vans", funded by the Home Office, and record it as one shared capability. Entered here as one record to avoid counting two shared vans as two deployments. The forces publish a running deployment register covering Redhill, Staines, Camberley, Ashford, Epsom, Woking and Guildford. The start year is deliberately left empty. Two readings of the same force register produced different dates for the first deployment at Redhill town centre — 13 November 2025 and 23 January 2026 — and the discrepancy was not resolved. The register itself should settle it before this field is filled. NEC Software is named by the forces in describing the Home Office-commissioned testing of the technology, which identifies the supplier but is not a separate procurement record.

Source

Surrey Police and Sussex Police live facial recognition pages, with published deployment registers · last checked 2026-09-04

Deployment

What we know

Two fully equipped LFR vans entered service on Monday 22 December 2025, first deployed on Cornmarket Street in Oxford city centre. The Police and Crime Commissioner announced the launch on 29 December 2025 and stated the vans would be used across the Thames Valley frequently. No supplier is named in the announcement, and no statutory authority or independent oversight body is identified. The Commissioner refers to working with the force to ensure safeguards are in place, and to trained operators, bespoke watchlists and automatic deletion of non-matching data, but those are descriptions of practice rather than an oversight arrangement, so the legal basis and oversight fields are left empty rather than filled from inference. The site's own news archive already carries reporting of subsequent deployments in Oxfordshire and Buckinghamshire, so the deployment was being covered before this record existed.

Source

Thames Valley Police and Crime Commissioner announcement, 29 December 2025 · last checked 2026-09-03

Deployment

Legal basis & oversight

Legal basis

No dedicated Canadian federal statute governing police facial recognition

Oversight

Not disclosed in available reporting

What we know

Toronto Police Service uses NEC's NeoFace Reveal to compare an image obtained during a criminal investigation against lawfully obtained booking photographs. Candidate images are reviewed by a trained facial recognition analyst before being supplied to investigators for further review. This entry was previously classified on this tracker as live facial recognition. That was wrong, and the correction matters beyond this single entry: retrospective image-to-mugshot searching and live scanning of passers-by are different technologies with different legal and civil liberties profiles, and conflating them inflates the apparent scale of live facial recognition deployment. The force's own published AI inventory identifies NeoFace Reveal as a high-risk AI technology, and Toronto operates an AI governance framework through the Toronto Police Services Board.

Source

Biometric Update · last checked 2026-08-27

Deployment

Legal basis & oversight

Legal basis

Statutory mandate on the Department of Homeland Security to build an integrated automated entry and exit system matching biographic and biometric records of noncitizens

Oversight

Department of Homeland Security Office of Inspector General

What we know

Captures a live photo of the traveller, taken by CBP or an approved airline, airport or cruise partner, and compares it against passport, US visa or other DHS-held photographs. A match verifies identity and the system flags people who may be inadmissible, including visa overstays. On non-match the traveller goes through traditional inspection, and US citizens and eligible foreign nationals may opt out and request a manual document check. CBP told Congress it had processed over 193 million travellers with facial biometrics, confirmed more than 163,000 visa overstays and prevented over 1,500 imposters from entering. The effectiveness evidence is mixed and both sides are recorded here. A DHS Office of Inspector General review found that between May 2019 and September 2021, CBP officers referred around 23,000 travellers aged 14 and over to secondary inspection when photos could not be matched. Separate reporting found CBP scanned over 23 million faces at more than 30 entry points during 2020 without identifying a single imposter.

Source

CBP congressional testimony; DHS Office of Inspector General report OIG-22-48 · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Not disclosed in the source; no dedicated statute specific to AI-based analytics identified

Oversight

US Government Accountability Office, which conducted the review this entry is sourced from

What we know

A GAO review found DHS law enforcement components using more than 20 categories of detection, observation and monitoring technology, including facial recognition and automatic licence plate readers alongside analytic software, some of which uses AI, plus access to third-party ALPR networks. GAO's own remit is bias-risk oversight, meaning this entry originates from an accountability review rather than a DHS disclosure. The review examined fiscal year 2023 usage; the report was published December 2024.

Source

US Government Accountability Office · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Sole-source federal contracts; no dedicated statute

Oversight

No dedicated regulator; Congressional oversight letter sent April 2026 by Rep. Dan Goldman and colleagues

What we know

Three related Palantir systems for ICE: (1) ICM, case management, $139.3m contract from 2022 renewed sole-source, combined deal reportedly exceeding $145m; (2) ImmigrationOS, a ~$30m no-bid 2025 contract modification providing 'near real-time visibility into instances of self-deportation' and managing the 'end to end immigration lifecycle'; (3) ELITE, an app generating deportation-target dossiers with a confidence score (out of 100) on a target's address. In M-J-M-A v Wamsley (D. Oregon), ICE officers confirmed under oath on 2 Dec 2025 that ELITE was used to identify arrest targets during Oregon raids driven by daily arrest quotas -- sworn testimony, stronger evidence than surrounding reporting. 404 Media reported ELITE drawing on HHS/Medicaid data; Palantir disputes this characterisation as misleading. A CMS-ICE data-sharing agreement exists; a preliminary injunction (Aug 2025) briefly blocked sharing in 20 states before a Dec 2025 ruling allowed some sharing to resume pending the case. Same vendor as this site's existing Palantir Gotham entry for German state police, where a Federal Constitutional Court judgment constrains automated analysis -- the US deployment has essentially the inverse legal position: no court constraint, sole-source contracts, and an operational target list rather than an investigative link chart.

Source

ACLU; American Immigration Council; court filings in M-J-M-A v Wamsley · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Arnold Ventures terms of use plus local court rules and statutes

Oversight

No formal regulator; ongoing independent validation including MDRC evaluation

What we know

An actuarial assessment using nine factors drawn from age and criminal history, explicitly excluding any interview, community ties, neighbourhood or marital status. It produces three estimates: failure to appear, new criminal arrest, and new violent criminal arrest while on pretrial release. Judges consider the scores alongside other information when setting release conditions, and scores are provided to the accused, the judicial officer, defence and prosecution. Piloted with selected jurisdictions from 2013 and released publicly in 2018 after five years of testing, built on around 750,000 cases from roughly 300 jurisdictions and validated on more than 500,000. Its deliberate exclusion of socio-economic and neighbourhood factors is a direct design response to bias criticism of earlier instruments.

Source

Advancing Pretrial Policy and Research; MDRC evaluation · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

State v Loomis (Wisconsin Supreme Court) held that a sentencing court may use a COMPAS risk assessment as a relevant factor but that it cannot be determinative, and required cautionary warnings about the instrument's limitations

Oversight

State court rules and case law; no dedicated regulator

What we know

A fourth-generation risk and needs instrument combining static criminal history with dynamic criminogenic factors such as social environment and employment, producing general recidivism, violent recidivism and pretrial misconduct risk scores. Designed for correctional case management and rehabilitation planning, its use expanded into judicial sentencing. Broward County, Florida is a widely cited example. The detailed input factor list circulating in secondary sources could not be confirmed against a primary vendor document, and is therefore not stated here as definitive. The legal position, by contrast, rests on the Loomis judgment and is well established.

Source

Harvard Law Review analysis of State v Loomis; California Courts summary · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Not detailed in the source reviewed

Oversight

Victorian Law Reform Commission has examined the deployment

What we know

Victoria Police uses iFace to search persons of interest against an offender image database. A Victorian Law Reform Commission report recorded 86 Victorian police stations with iFace capture points as of 2020, which is the most recent figure found; this entry should be treated as confirming the system exists and was operational as of that report, not as confirmation of its current 2026 scale.

Source

Victorian Law Reform Commission · last checked 2026-08-27

Deployment

Legal basis & oversight

Legal basis

None stated

Oversight

HMICFRS PEEL inspection regime referenced as context; no dedicated oversight body named

What we know

The first force in England and Wales to put the capability live, announced in March 2026. Records and transcribes control room calls inside a police-controlled environment, then auto-summarises and categorises them by topic, analyses repeat-caller patterns rather than raw call counts, identifies indicators of vulnerability not captured in operator notes, and checks whether crime prevention and forensic preservation advice was given. Outputs come with plain-language explanations and operational decision-making remains with trained staff. During pilot testing the force reported the system identified 21 percent more calls containing indicators of hidden vulnerability than its previous process. That is a force-reported pilot figure, not an independently replicated result. This finding was independently surfaced by two separate research passes, which is part of why it is recorded here rather than held.

Source

National Police Chiefs' Council · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Not specified in source; subject to an ICO compliance audit

Oversight

Information Commissioner's Office (consensual audit, executive summary published 18 August 2026); reasonable assurance rating for both live and retrospective use

What we know

West Yorkshire Police uses both live facial recognition and retrospective facial recognition. The Information Commissioner's Office audited the force between June 2025 and March 2026 as one of five in a consensual audit programme, publishing its executive summary on 18 August 2026. The audit gave West Yorkshire a rating of reasonable assurance of compliance for BOTH live and retrospective facial recognition. That is the stronger of the two outcomes among the audited forces on the retrospective side, where compliance was generally weaker across the programme. The audits consisted of policy and procedure reviews, interviews, records inspections, technology demonstrations and a virtual review of evidence documentation. Across all five forces the ICO made 107 recommendations covering compliance and best practice, every one of which was accepted or partially accepted. Recommendations applying to all forces include developing internal audit procedures, comprehensive logging of facial recognition use, and allowing sufficient time to carry out data protection impact assessments. Operationally, a deployment in Bradford city centre from 19 to 21 August 2026 identified four people wanted for criminal offences, all of whom were arrested. One was further detained on suspicion of possessing an offensive weapon after being found carrying a knuckle duster.

Source

ICO audit executive summary; Computer Weekly; Biometric Update · last checked 2026-08-28

Deployment

Legal basis & oversight

Legal basis

State “Strike Hard” campaign policy; described by Human Rights Watch as contrary to Chinese law and international human rights obligations

Oversight

None independent; documented externally through investigative research rather than official reporting

What we know

An algorithmic policing platform in Xinjiang that aggregates personal and behavioural data to generate alerts on individuals considered suspicious, which are transmitted to police for follow-up. Human Rights Watch, writing in April 2026, continues to describe IJOP in the present tense as a mass surveillance tool in place in the region. Labour transfer programmes tied to this surveillance apparatus have expanded from 11 to 14 provinces since a 2020 ASPI report first documented them, with new destinations including Shandong, Henan and Jiangsu. Uyghur returnees to China are reported to face facial recognition flagging at airports and immediate police interview, with documented cases of returnees disappearing within weeks of arrival. A note on a claim this entry does not make: material circulating about an IJOP 2.0 upgrade during 2026 traces to a low-quality source and is not recorded here pending primary corroboration. An earlier version of this entry referred to that upgrade; it has been removed. As with the Skynet entry, this rests on researchers working under significant state restriction on independent verification.

Source

Human Rights Watch, “China’s Algorithms of Repression: Reverse Engineering a Xinjiang Police Mass Surveillance App” · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

No dedicated Canadian federal statute governing police facial recognition. Forces state they rely on common law policing powers and, where available, court orders for camera footage

Oversight

Ontario Information and Privacy Commissioner consulted by York Police on best practice, per the force's own account

What we know

York Regional Police and Peel Regional Police use an Idemia facial recognition system to compare crime scene images against a combined database of booking photographs. Potential matches are reviewed by trained facial recognition analysts before being passed to investigators. The shared York and Peel database became operational in May 2024. This entry was previously classified on this tracker as live facial recognition, which was wrong for the same reason as the Toronto entry: this is investigative image-to-mugshot matching, not real-time scanning of people in public space. Kate Robertson of the University of Toronto's Citizen Lab has said she is not aware of any Canadian police service that in practice obtains prior judicial authorisation before using facial recognition, despite that being what she describes as the gold standard of privacy protection in criminal investigations.

Source

CBC News · last checked 2026-08-27

Deployment

Legal basis & oversight

Legal basis

Common law policing powers; DPA 2018 Part 3

Oversight

Force pilot evaluation, due to report after the pilot ends in November 2026

What we know

A time-limited trial running from February 2026 to November 2026, first deployed at London Bridge and extended to the London Underground with a deployment at Victoria station on 11 August 2026. The effectiveness evidence here is the most striking thing about this entry, and stands in sharp contrast to the Metropolitan Police's figures for the same technology family. Across 18 deployments between February and July 2026, British Transport Police logged more than 530,000 faces scanned, zero true matches, no arrests attributable to LFR, and one confirmed false alert at King's Cross on 24 February. The Met, by comparison, reported 173 arrests from roughly 470,000 faces at Croydon over a broadly comparable period. That divergence is worth recording plainly rather than resolving: the same class of technology, deployed by two forces in the same country in the same year, has produced radically different yields. Watchlist composition and deployment siting are the obvious candidate explanations, but neither force has published enough for that to be more than inference. The identification of the engine as NEC NeoFace M40 rests on secondary reporting rather than a BTP or NEC primary document.

Source

British Transport Police published information on its LFR pilot · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No dedicated statutory framework. A privacy impact assessment was submitted to Alberta's Office of the Information and Privacy Commissioner

Oversight

Edmonton Police Commission and the chief's committee, reviewing pilot results as of the time of writing

What we know

The first Canadian police service, and among the first anywhere, to trial facial recognition built into body-worn cameras rather than fixed or handheld devices. Axon supplied the camera hardware and Israeli firm Corsight AI the facial recognition model. The pilot ran through December 2025 and was extended roughly three weeks into January 2026 following a system outage. Officers' cameras ran the software in what EPS calls silent mode, checking faces against a watchlist of people with outstanding warrants for serious offences or flagged as high risk, reported variously as roughly 6,341 and 7,000 people, without alerting the wearing officer to a match; footage was reviewed afterwards by trained staff. Status note: the pilot itself concluded in January 2026 and results were with the Edmonton Police Commission and the chief's committee for review. This entry is deliberately held at trial rather than moved to discontinued, because a concluded pilot awaiting a decision is not the same as a technology that has been abandoned, and no decision either way has been published. Two independent research passes reached opposite conclusions on how to classify it, which is itself a reason to record the underlying facts and let the status follow a documented decision rather than an inference. A researcher at the University of Toronto's Citizen Lab described it as the most high-risk algorithmic surveillance program she had observed in Canada to date. Corsight AI's technology has reportedly been used for mass surveillance in Gaza, and the firm also supplies Essex Police in England. Axon's own AI ethics board recommended in 2019 that the company not pursue facial recognition on body cameras.

Source

CBC News (Freedom of Information documents), Associated Press, Biometric Update · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

None identified. Estonian police have stated publicly that no regulatory basis for facial recognition exists, and the separate plate-recognition camera network is not referred to in any statute.

Oversight

None identified. Estonian reporting notes the plate-recognition network has never been debated at Riigikogu level.

What we know

Estonia previously deployed facial-recognition-enabled cameras capable of watchlist alerts; that earlier deployment was reported as ineffective. The clearest recent confirmed use was a defined-area deployment during the Ukrainian president's 2024 visit. Estonia is also reported to be developing automated licence plate recognition and AI-based social media crime-prevention tools, though those are recorded here as context rather than as separately confirmed deployments. DIRECTLY CONTRADICTED, and recorded here rather than resolved. In May 2025 Roger Kumm, head of the Police and Border Guard Board's prevention and offence proceedings bureau, said publicly that the PPA cannot use facial recognition systems because no regulation for them exists. That statement is hard to reconcile with this entry's status of an ongoing trial, though it is not necessarily inconsistent with a one-off, time-limited deployment for a state visit. The status and start date on this entry should be treated as unconfirmed pending a source that settles which is right. What is well documented is adjacent but different. Estonian reporting establishes that the PPA operates a national automatic number plate recognition network: 250 cameras of which 228 were working, capturing more than 20 million vehicle images a month, concentrated in Harju, Tartu and Viljandi counties and typically sited at major junctions or settlement entrances. Cameras are often installed by agreement with, and paid for by, the local municipality. The PPA describes using it primarily in serious proceedings and to save lives. Reporting notes this network is not mentioned in any Estonian law and has never been the subject of parliamentary or public debate. Separately proposed, not operational: the Interior Ministry's development plans include a nationwide surveillance camera network supporting both face and plate recognition, with supporting analyses intended by 2027, and an AI capability for processing public-space video to detect threats. Politicians told ERR in March 2025 they had not been briefed on the plan, describing it as an official-level proposal.

Source

ERR (Estonian Public Broadcasting); Postimees; Eesti Ekspress · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

None stated

Oversight

None stated

What we know

Takes the caller's stated reason for calling 101, matches it with an AI model to a set of likely reasons, tells the caller how it understood them so they can accept or reject that reading, asks extra safeguarding questions for broad categories, and redirects non-police matters such as damage-only road traffic incidents, fly-tipping and noise complaints to councils, NHS 111 or the power-cut line before the caller joins a queue. If the model is in any doubt the call goes straight to police, and callers can always ask for a human handler. Use varies by force depending on existing telephone menus. A related programme covering AI transcription of 999 and 101 calls and linking of crime reports is backed by 16.5 million pounds. The government has stated it does not support police use of AI to answer 999 calls.

Source

GOV.UK, Home Office · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Ministry of Justice AI Action Plan for Justice; no dedicated statute identified

Oversight

None named

What we know

Digitally scans the contents of mobile phones seized from prisoners and applies AI language analysis to flag code words and messages indicating violence, escape plans or contraband smuggling for staff action. Trialled across the prison estate, having analysed over 8.6 million messages at the point of announcement in July 2025. Announced alongside a separate AI violence predictor, described as analysing factors including a prisoner's age and previous involvement in violent incidents in custody to predict individual risk and assess threat levels on wings. That second tool is recorded here as announced rather than deployed: the announcement does not state it has been trialled, and no deployment evidence has been found. Both rest on a single official source.

Source

GOV.UK, Ministry of Justice · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Common law policing powers; DPA 2018 Part 3. No dedicated statutory framework

Oversight

No dedicated oversight regime. Home Office consultation on facial recognition law closed February 2026 with results not yet published

What we know

Handheld devices allowing an officer to scan and identify a person on the spot. The Mayor of London confirmed in February 2026 that the Metropolitan Police was trialling the technology, following repeated questioning by Green Party London Assembly Member Zoe Garbett, and at a point when the force's own website stated it did not use operator-initiated facial recognition. Liberty Investigates reported that West Yorkshire Police documents anticipated deployment in 2026, though the force said no date had been set, and that Thames Valley had not confirmed timing. Dr Daragh Murray, who co-authored the 2019 independent review of the Metropolitan Police's live facial recognition trials with Professor Pete Fussey while both were at the University of Essex, has questioned whether handheld deployment can meet the same safeguards. Murray and Fussey jointly submitted a response to the Home Office facial recognition consultation in February 2026. An earlier version of this entry attributed that independent review to Queen Mary University of London, which was incorrect: the review was the Essex study. South Wales Police has separately trained over 500 officers to use a handheld facial recognition mobile app.

Source

Mayor of London disclosure, February 2026; Liberty Investigates · last checked 2026-08-20

Deployment

What we know

Established by Northamptonshire Police's own live facial recognition page and the deployment register it publishes, both read on 9 September 2026. The force's own wording is that it "is trialling the use of Live Facial Recognition (LFR) technology for the purposes of preventing and detecting crime and protecting the public", which is why the status is recorded as trial rather than operational. The published register lists deployments from 8 July 2023 onwards. The first six are at the British Grand Prix at Silverstone across 2023, 2024 and 2025, with watchlists ranging from 778 to 3,949 and no alerts recorded at any of them. Town centre deployments begin on 28 November 2025 at Abington Street and Fish Street in Northampton, with a watchlist of 3,317, eight alerts of which seven were true and one false, and three arrests. Further deployments follow in Kettering, Wellingborough, Corby, Northampton and Daventry, the most recent listed being 10 July 2026 at Foundry Place and Bowen Square in Daventry. The register was read twice in this run and gave the same first and last rows both times. Local reporting by the Northampton Chronicle & Echo on 10 December 2025 adds that seven arrests followed the first four town centre deployments and that the van used was loaned from Bedfordshire Police; that is secondary reporting, not a force document, and is recorded here as such. Vendor, legal basis and oversight are deliberately left empty. The force page links a legal mandate document, a data protection impact assessment, an equality and wellbeing impact assessment and a deployment policy, but states none of their contents on the page itself and names no supplier, and none of those documents was read in this run.

Source

Northamptonshire Police live facial recognition page and deployment register · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

Guidelines issued by Rigspolitiet and Koebenhavns Politi set out which case types the technology may be used in. Facial recognition is limited to serious crime against the person - the announcement names murder, attempted murder, aggravated violence and rape - to state security cases and to special operations. Object recognition may be used in all criminal cases.

Oversight

A data protection impact assessment was carried out and submitted to Datatilsynet, the Danish Data Protection Authority. Every use must be documented in the case file, and every video sequence the system identifies must be reviewed manually by an investigator before it is used in an investigation.

What we know

Established by the Danish National Police's own announcement of 7 March 2025, which is the source for every field above. The status is recorded as trial, not operational, because that announcement describes a pilot run by Rigspolitiet with Koebenhavns Politi, confined during the pilot to the Copenhagen police district and stated to run until autumn 2025. IMPORTANT CAVEAT FOR REVIEW: no source read in this run establishes what happened when that pilot period ended. Searching in Danish for an evaluation, an extension or a roll-out to other police districts returned nothing beyond the original announcement and its republication. The record therefore describes the last state established by evidence, and may understate current use. Danish reporting read in this run also refers to a preceding political agreement between the government and four parties permitting police facial recognition for serious crime against the person, state security cases and manhunts, and excluding real-time use; that agreement is not dated in the page read and is not relied on for any field here. Vendor left empty: no supplier is named in the announcement.

Source

Rigspolitiet · last checked 2026-09-06

Deployment

What we know

Recorded as a trial on the force's own wording, quoted verbatim from the Police Workplan Seminar 2026 page: 'At sea, the Police Coast Guard is trialling an Unmanned Surface Vessel (USV) to support maritime patrols and support incident response.' The status is the force's characterisation, not an inference. The same source states the vessel has been operating since March 2026, can operate fully autonomously in navigation, collision avoidance, anchoring and berthing, and has replaced one manned patrol twice a week. Sensors are given as a Seaborne Electro-Optics System, 360-degree stereoscopic cameras, warning systems and two-way communication. 'techSlug' is omitted deliberately. The autonomous-patrol-vehicles record was read in this run and is explicitly limited to road vehicles, distinguishing them from site-patrol robots and stating that they 'operate on roads, at vehicle scale'. A vessel does not belong under it, and forcing the link would misdescribe both. Fields left out: 'vendor', 'legalBasis' and 'oversight' are omitted because the source establishes none of them.

Source

Singapore Police Force · last checked 2026-09-07

Deployment

Legal basis & oversight

Legal basis

Lag (2026:806) om anvandning av AI-system for ansiktsigenkanning i realtid for brottsbekampande andamal, in force 1 July 2026. Passed by the Riksdag on 26 May 2026 following Prop. 2025/26:150 and committee report 2025/26:JuU28.

Oversight

Statutory. Use is subject to prior authorisation and to notification duties to affected individuals under sections 19-21 of the Act. The EU AI Act applies directly in Sweden alongside it.

What we know

Recorded with lower confidence than most entries on this tracker. The source describes a legal authorisation having been put in place, not a confirmed operational deployment; there is a meaningful difference between a law permitting a technology and evidence that it is actually being used, and only the former was confirmed in the material reviewed. Update, verified August 2026. The distinction this entry drew was right, and it has now been overtaken by events on both sides of it. On the law: the Riksdag approved the government's proposal on 26 May 2026, and Lag (2026:806) on the use of AI systems for real-time facial recognition for law enforcement purposes came into force on 1 July 2026. It permits the Police Authority, and in certain cases the Security Service, to use real-time facial recognition only where absolutely necessary, in three defined situations: to locate or identify a person suspected of being a victim of kidnapping, human trafficking or human exploitation, or who is missing and suspected of being a victim of crime; where there is an imminent risk the person will commit a serious offence endangering another's life or physical safety; or where a person is reasonably suspected of a serious offence carrying four years' imprisonment or more. The legislation is time-limited and subject to evaluation. On deployment: in June 2026 the Police Authority announced it had begun testing facial recognition systems, and stated that footage from the national camera platform, including images of individuals captured through general camera surveillance, would be used as the basis for testing the technology. Status is therefore updated from proposed to trial. This remains testing rather than confirmed operational use against live watchlists, and that distinction should be preserved on this entry until the force states otherwise.

Source

Sveriges riksdag; Polismyndigheten; Regeringskansliet (Prop. 2025/26:150) · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No national biometric surveillance framework exists in Australia. Western Australia's Privacy and Responsible Information Sharing Act 2024 is the first Australian law directly governing personal information in automated decision-making.

Oversight

State privacy watchdog was not consulted before the trial launched, according to advocacy groups

What we know

Western Australia Police became the first Australian law enforcement agency to use live facial recognition on the public when its five-month trial began on 22 June 2026, using NEC's NeoFace M40 from cameras mounted on marked police vans in Perth and Fremantle. Figures from the first week of the trial: more than 130,000 faces scanned, generating 33 alerts, leading to 18 arrests and two false alerts. Those numbers are worth reading in both directions. Two false alerts against 130,000 faces is a low error rate. It is also 130,000 people whose faces were converted to biometric templates in seven days so that 18 arrests could be made, and whether that ratio is proportionate is a policy question rather than a technical one that a more accurate algorithm resolves. Associate Professor Jake Goldenfein of Melbourne Law School has noted the established concern that these systems have historically performed less accurately for people of colour and for women, and least accurately for women of colour, because of the composition of the data the algorithms were trained on. WA Police has pointed to National Physical Laboratory testing of the NEC algorithm in selecting the technology. The trial has been criticised by privacy experts and legal advocates, and is being watched by other Australian forces as a precedent.

Source

ABC News; eWeek; Biometric Update · last checked 2026-08-28

Deployment

What we know

READ THE STATUS NOTE BELOW BEFORE APPLYING. The deployment is well established. Engineering News, carried by Polity on 28 July 2023, quoting Alderman JP Smith: acoustic triangulation and GPS pinpoint gunshots and relay the location to dispatch within seconds, with 'within a minute of a weapon being discharged, the police can be deployed' and information on shots fired and weapon type. Live in Hanover Park from December, Manenberg from February and Lavender Hill from March, with 35 firearms and 400 rounds confiscated and 50 arrests, and 10 million rand budgeted for a Nyanga expansion in 2023/24. Cape Argus, 18 June 2026, and IOL, 19 June 2026: the system began as a pilot in Hanover Park in 2016 and expanded to Manenberg and Lavender Hill from 2023; the City invested about 860 million rand over three years across CCTV, drones, body cameras, dashcams, aerial surveillance and gunshot detection; the City described the technology as useful for alerting staff to active shooting incidents and for identifying hotspots to guide deployment. Status note. The current status is genuinely unresolved and 'paused' is a deliberately cautious reading, not a confident one. What is quoted in June 2026 is Smith saying: 'The three-year gunshot detection technology tender has run its course and is currently under review.' That establishes the contract ended; it does not on its own establish that the programme is over. Separately, Police.AI's own archive carries an Inside Metros story of 5 August 2026 whose summary states that Cape Town's city council ended the use of the system in mid-2026, but that article could not be fetched in this run and the site's own summary is not treated here as a primary source. If you can open that article and it confirms the council ended the deployment, change this record to discontinued. Nothing read establishes that the system is still running, which is why it is not proposed as operational. Effectiveness is contested and the record does not take a side: the City calls the technology useful, while crime-focused organisations quoted in the same reporting question whether it achieved its objectives given continued high violence. Fields left out and why. 'legalBasis' and 'oversight' are omitted: nothing established. Cost is not given as a field because the 860 million rand figure covers a basket of technologies, not this system.

Source

Cape Argus · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

Ley 5688 (Sistema Integral de Seguridad Pública de la Ciudad), under which the system was permitted to search only for persons listed in the national fugitives register CONARC or subject to a judicial order. Judicially suspended on 21 April 2022; its implementation was declared unconstitutional on 7 September 2022 and that ruling was upheld on appeal on 29 April 2023.

Oversight

The city legislature approved the system subject to control by an ad hoc commission of legislators and by the Defensoría del Pueblo. CELS reports that as of February 2024 the court required, before any resumption, an effective audit by the control authorities, a budgetary framework for it, access to the supplier's software information, and an evaluation of possible discriminatory bias.

What we know

Operated in the Buenos Aires underground and public space from 2019, matching faces against CONARC. Judge Roberto Gallardo suspended it on 21 April 2022 after an amparo brought by the Observatorio de Derecho Informático Argentino (ODIA); raids on city offices found more than 9 million queries against the national identity register when CONARC held only some 35,000 to 40,000 records, including searches on the President, the Vice-President, judges, journalists, activists and business figures, and on more than 15,000 people not on the CONARC list. The city said these were identity validations for permits, pandemic procedures and emergency services. On 7 September 2022 Judge Elena Liberatori declared the implementation unconstitutional, finding the system had been launched and run outside the existing regulatory framework and without control mechanisms; the Cámara en lo Contencioso Administrativo rejected the city's appeal on 29 April 2023. Status is 'paused' rather than 'discontinued' because CELS, Fundación Vía Libre and ODIA all report the system as suspended, with conditions specified for resumption, rather than terminated. vendor is left empty: the AlSur regional report of June 2025 names Danaide as the supplier, but that is a single secondary source and no primary document read here confirms it. Sources in this record are in Spanish. Note that the city's prosecution service separately acquired Clearview AI in June 2025 — a distinct deployment with its own record.

Source

CELS · last checked 2026-09-10

Deployment

Legal basis & oversight

Legal basis

New Orleans city ordinance restricting police facial recognition; the ACLU alleges the department's use breached it

Oversight

New Orleans City Council; ACLU of Louisiana investigation

What we know

Project NOLA is a private non-profit founded in 2009 by a former police officer, based on the University of New Orleans campus, acting as a hub for live video from thousands of cameras on private volunteer properties whose owners pay annual connectivity fees. Around 200 of the most sophisticated cameras run live facial recognition, continuously matching passers-by against hot lists of roughly 250 individuals. An automated voice alerts staff on a match; staff review and relay information to police once an evidentiary threshold is judged met, including unsolicited tips. From early 2023 until April 2025 the police department received real-time facial recognition alerts from a network of over 5,000 cameras, more than 200 facial-recognition equipped, despite city restrictions. The department suspended the real-time alert partnership pending a legal compliance review; the camera network was later reopened to police while facial recognition alerts stayed suspended. The City Council has moved towards an ordinance that would make New Orleans the first US city to formally authorise this form of facial recognition policing. Recorded as paused to reflect that the camera access continues while the facial recognition alerting does not.

Source

NPR; ACLU of Louisiana · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Internal force policy; no dedicated statutory framework for generative AI use

Oversight

House of Commons Home Affairs Committee inquiry

What we know

West Midlands Police switched off force-wide access to Microsoft Copilot in January 2026 after the tool generated false intelligence about a West Ham versus Maccabi Tel Aviv fixture, and associated disorder, that had never taken place. The fabricated information reached briefing material and shaped a risk assessment. Recorded as paused rather than discontinued. Acting Chief Constable Scott Green described the shutdown as temporary, in place until further notice, both in January and again in March 2026. No evidence of reinstatement had emerged as of August 2026, but a force describing a suspension as temporary is not the same as a decision to stop using a technology permanently, and this tracker records the difference. The force initially denied that AI had generated the erroneous intelligence; the Chief Constable subsequently corrected that evidence. The Home Affairs Committee reported in February 2026 that West Midlands Police failed to apply proper due diligence to false information generated by AI, and criticised what it called a remarkable lack of professional curiosity. This pause did not extend nationally. A Sky News survey published in February 2026 found at least 21 English forces still using Microsoft Copilot.

Source

House of Commons Home Affairs Committee correspondence, reported by ITV News · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Garda Siochana (Recording Devices) Act 2023, as proposed to be amended by the Garda Siochana (Recording Devices) (Amendment) Bill 2025

Oversight

Subject to ongoing legislative process

What we know

Irish government documents describe proposed retrospective facial recognition technology for serious crime, including searching CCTV, intended to automate searches that would otherwise require manually reviewing very large volumes of footage. Live facial recognition was separately reported as under consideration for terrorism, national security and missing-persons cases, subject to stated safeguards. No operational deployment confirmed at the time of the source reviewed. Update, verified August 2026: the legislation has advanced materially. The General Scheme was approved and published in December 2023 and referred for pre-legislative scrutiny; the Justice Committee reported in February 2024. On 11 December 2025 the Minister for Justice secured cabinet approval to publish the Garda Siochana (Recording Devices) (Amendment) Bill 2025, which would allow An Garda Siochana to sort, filter and compare images from CCTV, mobile phone footage and evidence submitted by the public, for suspected arrestable offences, missing person searches and State security. Safeguards including advance written authorisation by a Chief Superintendent were set out in the General Scheme, with detail deferred to a Code of Practice to be drafted by An Garda Siochana and laid before the Oireachtas. Contested: the Irish Council for Civil Liberties and Digital Rights Ireland argue in a joint briefing that the Bill introduces a new term, 'biometric analysis', which overlaps with the EU AI Act's definitions of biometric categorisation and biometric identification without adopting the Act's terminology or safeguards. The Irish Human Rights and Equality Commission separately described the proposals as a radical upturn in State policing powers and called for safeguards written into the statute rather than left to a future Code of Practice. Still not operational: this remains proposed legislation, not a deployed system.

Source

Department of Justice (Ireland) press release; Oireachtas Library & Research Service · last checked 2026-08-27

Deployment

Legal basis & oversight

Oversight

fedpol states that matches produced automatically are verified manually by its own facial image experts, that comparison is carried out exclusively against images already stored in AFIS from police person-queries - social media images and identity document photographs are excluded - and that use is confined to individual cases after an offence, with real-time surveillance ruled out.

What we know

Established by fedpol's own page on Gesichtsbildabgleich, which describes the automated fingerprint identification system AFIS being renewed and extended with a facial image comparison module, under which facial image traces from crime scenes are compared against stored facial images of known persons. The page refers to the project as AFIS2026 and gives a contract award announced on 10 March 2025. Status is recorded as proposed rather than trial or operational because the page describes a system being built and nothing read in this run establishes that the module is in use. Legal basis left empty: the fedpol page read does not state which law and article authorise the module, and Swiss commentary read in searching for it disputes whether the existing basis is adequate - that dispute is recorded in the accompanying report rather than asserted here. Vendor left empty: the contract award is mentioned but the awardee is not named on the page read. The 'since' field is left empty for the same reason - a contract award date is not a deployment date.

Source

fedpol (Swiss Federal Office of Police) · last checked 2026-09-06

Deployment

Legal basis & oversight

Legal basis

The Italian Data Protection Authority (Garante) examined the system as designed and found it unlawful for lack of an adequate legal basis and the potential for blanket surveillance

Oversight

Italian Data Protection Authority (Garante per la protezione dei dati personali)

What we know

SARI Real Time was designed to analyse live camera feeds against a predefined watchlist of up to 10,000 faces and alert operators on a match. The Garante's review found it was not yet active and ruled it unlawful as designed. This is exactly the kind of regulatory finding this tracker exists to document: a live facial recognition system blocked by a data protection authority before deployment, not after.

Source

Garante per la protezione dei dati personali, decision of 16 April 2021 · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

None yet; described as preliminary discussion

Oversight

Malta's Data Protection Authority, which is the source recording these discussions

What we know

Recorded with low confidence deliberately. Malta's Data Protection Authority states law enforcement authorities began preliminary discussions in 2025 about deploying remote biometric identification at an airport. This is an early-stage exploratory discussion, not a confirmed system, procurement, or trial, and should not be read as evidence anything has actually been deployed. Sourcing status, checked August 2026: no citable public document beyond the data protection authority's own reference to preliminary discussions was located during a dedicated search. No procurement notice, tender, vendor, trial, ministerial statement or parliamentary record was found. That absence is itself the accurate position and is recorded here rather than papered over with a link to secondary coverage that would only restate the same single underlying claim. This entry should continue to be read as an early-stage discussion on the public record, not as evidence that Malta has deployed or is close to deploying any biometric identification system.

Source

Malta Data Protection Authority (no direct link on file) · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Data Sharing Agreement between MoJ and GMP; no deployment authorisation identified

Oversight

None external; MoJ confirms research stage only

What we know

Originally 'Homicide Prediction Project', renamed 'Sharing Data to Improve Risk Assessment'. Began Jan 2023, using Ministry of Justice data, Police National Computer records and GMP data to build models identifying people at greatest predicted risk of committing homicide. The MoJ-GMP Data Sharing Agreement covers 100,000-500,000 people, including names, dates of birth, ethnicity and gender. A project timeline obtained by Statewatch shows the project completed Dec 2024 but not deployed; MoJ confirms research stage. Disputed: Statewatch says sensitive data including domestic abuse and self-harm information was used, covering people never convicted of any offence; officials deny this, stating only convicted persons' data was used -- both positions recorded pending resolution. Connects to this site's existing OASys entry via an MoJ finding that OASys's predictive validity was 'greater for white offenders than offenders of Asian, Black and Mixed ethnicity'. Recorded as proposed rather than operational: developed but explicitly not deployed.

Source

Statewatch (FOI documents); The Register · last checked 2026-08-24

Deployment

What we know

Established by two independent sources reporting the same statement by Interior Cabinet Secretary Kipchumba Murkomen. Capital FM, 24 June 2026: the system is in the final stages of procurement, estimated at 22 to 25 billion Kenyan shillings, to be rolled out across six major cities, connecting surveillance cameras to national identity databases through a central command system, with deployment expected within three to six months, and phased implementation possible through private sector partnerships. Capital FM reports the stated purpose as identifying and tracking individuals accused of infiltrating protests and engaging in violence including property destruction. Biometric Update, June 2026: the same announcement, naming the six cities, the roughly 25 billion shilling cost, and integration of the National Registration Bureau database with police CCTV so faces can be matched against civil ID photographs; Murkomen is quoted saying 'Once we establish this technology, we will slowly build in the data of various people based on information in the ID', and defending the plan on the ground that cameras would be installed only in public spaces. Status is proposed, and deliberately so. A procurement in its final stages is not a deployment. Nothing read establishes that any camera under this programme has been installed or that any facial matching against the identity register has occurred. The three-to-six month timeline from June 2026 means this record should be re-checked rather than left to the next cycle. The stated purpose is unusually explicit and is recorded here because it matters to how the system should be read: this is a national biometric identification network whose announced first application is the identification of protesters. Fields left out and why. 'vendor' is omitted: no supplier is named. 'legalBasis' is omitted: no statute, regulation or authorisation instrument was established by any page read, and populating a police facial recognition database from the civil identity register is precisely the point on which a legal basis would be needed. 'oversight' is omitted: none established. 'since' is omitted: nothing has started. Second claim, deliberately not made a record. Biometric Update also states that Kenya's National Police Service already operates NEC facial recognition within the Integrated Command, Control and Communication Centre launched in 2018. That rests on that single article, no corroborating source was found, and it is left on the watch list rather than asserted.

Source

Capital FM · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

None yet; under consideration

Oversight

Scottish Biometrics Commissioner; Scottish Police Authority

What we know

The Scottish Police Authority launched a National Conversation in 2024 on the potential use of live facial recognition by Police Scotland. The ICO has been engaging with Police Scotland and the SPA throughout and will continue to input as proposals develop, and separately responded in June 2026 to a Scottish Government consultation reviewing the Scottish Biometrics Commissioner's functions. The ICO's own August 2026 audit statement covers England and Wales 'where LFR is currently in use', which is a citable statement that LFR is not currently operational in Scotland. Scotland is unusual among tracker entries in having a named biometrics-specific oversight body already in statute.

Source

ICO, 18 August 2026 statement · last checked 2026-08-24

Deployment

Legal basis & oversight

Legal basis

None yet; the ICO states it is 'feeding into the Northern Ireland Justice Bill'

Oversight

None named yet

What we know

No confirmed operational facial recognition use in Northern Ireland. The ICO's August 2026 statement says it is feeding into the Northern Ireland Justice Bill, and separately says its audit findings will be shared with the PSNI alongside Police Scotland. Recorded as proposed on the strength of the Justice Bill reference; the underlying legislative content was not established in this research pass.

Source

ICO, 18 August 2026 statement · last checked 2026-08-24

Deployment

Legal basis & oversight

Legal basis

None specific. The ministry states the new team's principal task is to develop safe and transparent rules for AI use, consistent with European data protection and civil rights standards.

Oversight

A ministerial AI team, established 22 April 2026, is tasked with setting the rules. No independent regulator identified; reporting notes the announcement carried no implementation timetable, costs, or stated mechanism for auditing the algorithms.

What we know

Recorded with low confidence on specifics deliberately. Poland's Interior Ministry announced in 2026 that AI would support criminal investigations, evidence analysis, searching large volumes of recordings, pattern detection, connecting related facts across cases, emergency-call analysis and facial recognition for border protection. This is a programme announcement of intent and direction; no specific named system, vendor or confirmed operational deployment was found in the material reviewed, and this entry should be read as a statement of policy direction rather than a confirmed technology in use. Sourcing and detail, verified August 2026. On 22 April 2026 the Interior Minister, Marcin Kierwinski, announced the creation of a dedicated AI team covering the Police, Border Guard and the 112 emergency notification system. The ministry's own statement describes intended uses as: analysis of evidential material, with AI able to search thousands of hours of recordings or thousands of documents; detection of patterns linking apparently unconnected facts; and automated detection of anomalies for cybercrime. The Border Guard is to be supported by automation covering biometrics and image recognition, including automatic detection of border violations. The minister separately stated AI would be used for facial image recognition in border protection. Still a programme announcement rather than a deployment. The ministry gives the team's principal task as developing safe and transparent rules for AI use. Polish reporting notes the announcement contained no implementation timetable, no indication of which systems would go live first, no costs, and no stated mechanism for scrutinising how the algorithms operate. The minister's stated starting point was that the existing IT tools for data processing were built more than twenty years ago. The team held its first meeting in May 2026.

Source

Ministerstwo Spraw Wewnetrznych i Administracji (MSWiA), official announcement · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Proposed amendments to the Police Register Act (politiregisterloven) and the Police Register Regulation, issued for public consultation by the Ministry of Justice and Public Security on 21 November 2025 with a deadline of 16 January 2026.

Oversight

The proposal would require prior approval from the prosecution authority (paatalemyndigheten) for targeted searches against suspected or convicted persons, exempting initial identification of an unknown suspect; would require each use to be documented; and would introduce annual reporting to Datatilsynet. The stated aim is protection equivalent to Article 26(10) of the EU AI Regulation.

What we know

Established by the Norwegian Ministry of Justice and Public Security's own consultation documents, read twice in this run - the announcement page and the consultation document page - which agree on the ministry, the date, the deadline and the substance. The proposal covers retrospective biometric remote identification only, described as automated comparison of collected biometric data against corresponding data in a register, for example matching images from video surveillance against stored photographs. The stated purposes are identifying wanted or suspected persons in cases such as violence and sexual offences, and identifying victims and witnesses. The ministry's announcement states that Norwegian police do not today use real-time biometric remote identification systems. Status is proposed, not trial: nothing read establishes that any system has been procured or piloted. IMPORTANT CAVEAT FOR REVIEW: the consultation closed on 16 January 2026, roughly eight months before this run, and no source read establishes what has happened since - whether a bill has been laid before the Storting, amended or dropped. The record should be re-checked before it is relied on as current, and the status may already be out of date in the proposed-to-trial direction.

Source

Norwegian Ministry of Justice and Public Security · last checked 2026-09-06

Deployment

Legal basis & oversight

Legal basis

Privacy Act and RCMP internal security protocols govern the test data itself; no legislative authorisation yet exists for the capability this is evaluating.

Oversight

Internal RCMP process as described in its own departmental plan; no external oversight body specified

What we know

The RCMP's 2026-27 Departmental Plan states it will collaborate with multiple vendors to review booking photographs collected from Canadian law enforcement agencies, testing algorithms for quality, performance, accuracy and demographic bias in a Canadian context, towards a potential future national facial recognition capability. This is evaluation of algorithms against historical data, not a deployment against the public, which is why it is recorded here as proposed rather than trial. Verified directly against the RCMP's own published departmental plan.

Source

Royal Canadian Mounted Police, 2026-27 Departmental Plan · last checked 2026-08-18

Deployment

Legal basis & oversight

Legal basis

Privacy Act 1988; Australian Government Agencies Privacy Code

Oversight

Office of the Australian Information Commissioner (OAIC), joint investigation with UK ICO

What we know

This trial was run specifically by the Australian Federal Police's Australian Centre to Counter Child Exploitation (ACCCE), not the AFP generally. The AFP's own account, given in a Senate Estimates brief, states ACCCE personnel ran a limited pilot from November 2019 to January 2020, testing whether Clearview could match side-profile photographs of possible child exploitation victims against frontal reference images; temporary accounts were terminated in March 2020. The Australian Information Commissioner subsequently found the AFP had failed to comply with its privacy obligations, including a requirement to complete a privacy impact assessment before the trial. Separately, and consistent with the free-trial period documented for other Australian police services in this entry, Queensland, Victoria and South Australia police also had access during the same general window.

Source

Office of the Australian Information Commissioner, Commissioner-initiated investigation determination · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Not established in available sources

Oversight

Reviewed by Social Finance, which could not retest the models

What we know

A database launched in 2016 holding records on close to half a million Bristol residents, with child sexual exploitation and child criminal exploitation risk models flagging children at risk; the child criminal exploitation model was introduced in 2019. Stopped around June 2023 after council staff judged the models not fit for operational use, with recent victims of sexual offences scoring below burglary perpetrators. No records of the decision were retained. Reviewers from Social Finance could not retest the models because the source code and model variables could not be found. That combination, a discontinued predictive system with no decision record and no recoverable model, is itself the finding worth recording.

Source

Liberty Investigates · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Custody decision support

Oversight

Academic review; extensive published critique

What we know

An early and unusually well-documented custody risk-scoring tool, developed by Durham Constabulary with the University of Cambridge, using random forest models trained on historical criminal justice data to sort individuals into risk categories. Operational from 2016 until 2021, when Durham Constabulary discontinued it. The stated reason was the resources required to continually refine and refresh the model to satisfy ethical, legal and governance requirements. A Freedom of Information response obtained by Fair Trials showed the tool assessed 12,200 people a total of 22,265 times across that period. Its use of a commercial postcode-based variable attracted particular criticism. The variable was Experian's Mosaic socio-geodemographic segmentation product, a commercial consumer classification dataset rather than a policing one.

Source

Published academic evaluation and force statements · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

No dedicated statute; registered in the Dutch national algorithm register

Oversight

Dutch national algorithm register provides public transparency; no dedicated regulator identified

What we know

Divided cities into small grid cells and combined police crime records with national statistics including average neighbourhood income, benefit recipients, household composition and distance to the nearest known offender, outputting hotspots and hot times in four-hour blocks. Officers were sent to drive through flagged areas, often for only a few minutes. The model was closed, so officers could not see why a cell was flagged. An indicator counting non-Western immigrants was removed in 2017 after being found to add no predictive value. Piloted in 2015 in Enschede, Groningen-Noord, Hoefkade and Hoorn, then rolled out from 2017 across more than half of the then 167 police departments. Halted in January 2026 and discontinued after an internal report concluded its predictive capability was extremely limited and the force could never demonstrate its effect.

Source

AlgorithmWatch; Dutch national algorithm register · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Hellenic DPA ruled the procurement, operation and use unlawful in Decision 45/2025 (31 Dec 2025)

Oversight

Hellenic Data Protection Authority

What we know

1,000 handheld devices for Hellenic Police performing facial recognition, fingerprint ID, and document/plate scanning during street stops, matching against national, EU and international databases including SIS II, VIS, EURODAC and Interpol-linked systems. Contract signed 2019 with Intracom Telecom, €4m, 75% funded by the EU's Internal Security Fund. Devices were delivered fully functional. Following a complaint by Homo Digitalis (Mar 2020), the Hellenic DPA ruled the whole programme unlawful in Decision 45/2025, dated 31 Dec 2025 -- an EU-funded, fully procured and delivered biometric system found unlawful after the money was spent. Reported as targeted at stops of undocumented migrants, placing it in borders/immigration as much as ordinary policing. Current operational status unclear following the ruling -- recorded as discontinued pending confirmation. Also in Greece: the Ministry of Asylum and Migration was fined a record €175,000 by the Hellenic DPA over the KENTAUROS and HYPERION border/camp management systems.

Source

EDRi; Homo Digitalis; Hellenic DPA Decision 45/2025 · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

Force procurement; no dedicated statute

Oversight

None identified beyond the force itself

What we know

The first UK force to test PredPol. The force cancelled the contract after saying it was challenging to demonstrate that the programme reduced crime. This is the same pattern seen in Los Angeles and in the Dutch Crime Anticipation System: the operating agency itself was unable to show the software worked.

Source

Liberty Investigates · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Municipal contracting; no dedicated statute

Oversight

LAPD Office of the Inspector General reviewed the deployment

What we know

The LAPD was the earliest adopter and co-developer, dropping the system in 2020 citing budget, after an inspector general report found it unclear whether the software affected crime trends. Geolitica has since ceased operations, with SoundThinking acquiring specific intellectual property and hiring the engineering team. A Gizmodo and Markup analysis of millions of predictions found on an unsecured server reported the software disproportionately and often relentlessly targeted low-income communities of colour.

Source

WIRED; Gizmodo and The Markup analysis · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

City council authorisation

Oversight

City council vote following public disclosure

What we know

Police Chief Mike Canfield disclosed to the city council in February 2026 that federal agencies had gained access to the city’s Flock camera data without the city’s knowledge. The council voted unanimously to terminate the contract and the department discontinued use. Included as a specific, well-documented instance of a US agency withdrawing a deployment after an oversight failure, distinct from the wider Flock Safety entry tracked separately.

Source

City council disclosure, reported by State of Surveillance and the ACLU · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

None in place at the time; this absence is what the regulatory reprimand below addresses

Oversight

Finnish Data Protection Ombudsman, who reprimanded the National Police Board

What we know

Finland's National Bureau of Investigation ran approximately 120 Clearview AI searches in early 2020 while trying to identify possible victims in child sexual abuse material, producing one hit, and separately tested Arachnid for the same purpose; police concluded Clearview was unsuitable for the task. The Finnish Data Protection Ombudsman subsequently reprimanded the National Police Board because the trial had taken place without appropriate controller approval or supervision.

Source

Office of the Data Protection Ombudsman (Finland), statutory reprimand · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

NSW Police Force internal policy and guidelines rather than statute

Oversight

NSW Law Enforcement Conduct Commission (Operation Tepito); NSW Bureau of Crime Statistics and Research evaluation

What we know

Chimera collated police information sources weighted by perceived reliability and computed a Crime Severity Index score from charges and offending indicators, scoring each offence against the Australian Bureau of Statistics National Offence Index and typically covering the prior six months. A benchmark score limited the candidate list. Intelligence officers reviewed the shortlist alongside offending history and local crime patterns, could add people from local knowledge, documented a justification, and a senior officer approved selection. Those selected were subjected to increased overt and covert police attention. STMP was introduced in 2000, with STMP III and Chimera from November 2020. It was discontinued for children and young people in October 2023 and for adults from December 2023, following the Law Enforcement Conduct Commission's Operation Tepito investigation.

Source

NSW Law Enforcement Conduct Commission, Operation Tepito final report · last checked 2026-08-20

Deployment

Legal basis & oversight

Legal basis

Municipal contracting

Oversight

None identified

What we know

A self-driving camera-equipped patrol robot deployed in Times Square subway station, accompanied by police officers, as a trial of crime prevention effectiveness. Leased at 9 dollars an hour under a seven-month contract worth 12,250 dollars. The NYPD retired the robot in January after four months deployed and confirmed the pilot was complete. Huntington Park Police Department in California renewed a Knightscope contract for a sixth year in August 2024, though that detail is vendor-sourced and carries lower confidence.

Source

Fox News; Knightscope contract announcement · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

City anti-violence initiative; commuters and civil rights groups argued mass subway scanning is impractical and unconstitutional

Oversight

Evolv has faced federal investigations into its marketing by the Federal Trade Commission and the Securities and Exchange Commission

What we know

A month-long pilot across 20 subway stations, beginning at Fulton Street on 26 July, reporting 2,749 scans. It found zero firearms, produced false positives and detected some knives. The department did not disclose how many passengers opted out or how many officers were needed to run the devices, and no agreement with Evolv had been finalised while the pilot outcomes were being assessed. Recorded as discontinued on the basis that the pilot completed with no continuing deployment established. LA Metro has run a comparable pilot at Union Station.

Source

Associated Press · last checked 2026-08-26

Deployment

Legal basis & oversight

Oversight

Reviewed by the Independent Police Conduct Authority, which reported on 30 November 2020.

What we know

Established by the Independent Police Conduct Authority's published outcome of 30 November 2020, an official oversight body's own record. The Authority found that New Zealand Police trialled Clearview AI facial recognition software between February and March 2020; that the trial was approved by an internal governance group but undertaken without approval from the Police Executive; and that there was no consultation with the Privacy Commissioner or the Government Chief Data Steward. The software 'was deemed not operationally viable'. Police subsequently established a central governance body for emerging technologies and released a policy in September 2020 governing how emergent technologies may be tested or trialled. The Authority's broader assessment was that Police use of surveillance technologies was generally conservative, well-considered and broadly compliant with privacy and data security principles. Status is discontinued rather than paused: the Authority records that the software was found not operationally viable and that the response was a new governance regime, not a suspension pending resumption. Nothing read in this run suggests any later Clearview AI use by New Zealand Police. This record is included because the withdrawal is the substance: it is the event that produced New Zealand Police's emergent technologies governance framework, the Expert Panel and the later facial recognition policy, all of which the two other New Zealand records in this changeset depend on. Fields left out and why. 'legalBasis' is omitted - the point of the finding is that the trial proceeded without the approvals it should have had. The number of images searched is not stated in the Authority's published outcome and is not asserted anywhere in this record.

Source

Independent Police Conduct Authority · last checked 2026-09-09

Deployment

Legal basis & oversight

Legal basis

Privacy Act (federal public sector); found in breach

Oversight

Office of the Privacy Commissioner of Canada, Special Report to Parliament

What we know

The RCMP's use of Clearview AI ended after the Office of the Privacy Commissioner found in 2021 that it violated the federal Privacy Act, since Clearview's underlying image collection was itself unlawful. Clearview ceased Canadian operations shortly afterward. In response, the RCMP established the National Technology Onboarding Program (NTOP) in 2021 to centralise assessment of new operational technologies. The RCMP's own 2026 transparency publication states it currently uses only 'face matching' (grouping and sorting lawfully obtained images and video, not searching to identify an unknown individual) rather than full facial identification, and that it does not use facial recognition on footage captured by its drone fleet. Separately, other Canadian police services have moved into live and image-search facial recognition; see the Edmonton, York/Peel and Toronto entries.

Source

Office of the Privacy Commissioner of Canada, Special Report to Parliament · last checked 2026-08-26

Deployment

Legal basis & oversight

Legal basis

No dedicated statutory basis. The EU AI Act, in force from July 2024, classified tools of this kind as high risk, which is widely reported as a factor in the decision to stop using it.

Oversight

Spain's Consejo de Transparencia y Buen Gobierno, which ordered the Interior Ministry to release information about the tool following requests by the transparency organisation Civio

What we know

VeriPol applied natural language analysis to written robbery reports to flag characteristics statistically associated with false reports. It analysed the text of a report a victim or witness gave rather than physical evidence or camera footage, which makes it a materially different form of police AI from the video and biometric systems that make up most of this tracker. Launched nationally in October 2018 after a 2017 pilot in Malaga and Murcia, and promoted at the time as over 90 percent accurate. Spain's Dirección General de la Policía confirmed to the transparency organisation Civio that the tool stopped being used on 21 October 2024, giving as its reason that it lacked validity in judicial proceedings. That date is three months after the EU AI Act came into force. The evidence base was thin from the outset and this is central to understanding the entry rather than incidental to it. The model was trained on 1,122 robbery reports from 2015, and rested on a starting assumption that 57 percent of violent robbery reports are false. A University of Valencia analysis identified serious design deficiencies. Usage fell sharply over time, from roughly 84,000 reports analysed up to October 2020 to 3,762 in 2022. The Interior Ministry has declined to release which police stations deployed it or how often it was used, and a transparency complaint on that point remains open.

Source

Civio, reporting the Directorate General of Police's confirmation · last checked 2026-08-27