United States, New York City

Metropolitan Transportation Authority (New York)

AI video analytics for weapons, unattended items and unsafe behaviour on subway and bus cameras

ProposedOfficial publicationPAI-DBB45D05

Body
Metropolitan Transportation Authority (New York)
Country
United States
Area
New York City
Technology
AI video analytics for weapons, unattended items and unsafe behaviour on subway and bus cameras
Supplier
Not stated in the source
Status
Proposed
Planned by a named body: a procurement, approved budget, announced pilot or consultation. Not yet in use.
Stage
On the horizon
Since
Not known
Legal basis
Not stated in the source
Oversight
Not stated in the source
Evidence
Official publication
Police force, government department or public body publication
Source
MTA, Request for Information 0000532398, 5 December 2025
Last checked
Record code
PAI-DBB45D05

Notes

The MTA issued Request for Information No. 0000532398, 'AI Video Analytics for Forbidden Objects, Unattended Items, Unsafe Behaviors', dated 5 December 2025, with responses due 30 December 2025. The document says the MTA is gathering information to inform a possible 'Proof of Value' for a scalable system that analyses video from cameras on subway cars and buses, covering identification of weapons, firearms and other hazardous materials, unattended items, and systems that 'Detect and predict unsafe or anomalous behaviors', including high-density crowd surges. It notes that more than 15,000 cameras across about 472 subway stations are monitored manually. It states that the RFI does not constitute a solicitation or a commitment and that no contract will be awarded from it. An earlier version of the RFI on the MTA site carries a date of 4 December 2025 and earlier deadlines. The RFI names no vendor, pilot or budget and does not say who would receive alerts. GovTech reported on 2 May 2025 that the MTA's chief security officer, Michael Kemper, said AI software could eventually help the agency and law enforcement spot problematic behaviour in real time, with alerts going to security guards or police officers, and that facial recognition would reportedly not be part of the effort. Listed as proposed on the basis of this exploratory step; the MTA document does not show a decision to procure.

AI video analytics for weapons, unattended items and unsafe behaviour on subway and bus cameras elsewhere in United States

See them in the tracker
AI video analytics for weapons, unattended items and unsafe behaviour on subway and bus cameras elsewhere in United States
DeploymentCountrySupplierSinceStatusChecked
US Customs and Border ProtectionAutonomous Surveillance Towers United StatesBorder Patrol sectors on the southwest and northern borders Anduril Industries 2018 Operational
US Customs and Border ProtectionMatroid automated item of interest detection in Border Patrol imagery United StatesBetween ports of entry Matroid 2020 Operational
US Customs and Border ProtectionVessel detection from camera imagery United StatesMaritime border areas JHU APL 2025 Operational
US Customs and Border ProtectionCBP Common AI Service: image analysis and data correlation United States — — Proposed
US Customs and Border ProtectionTethered Aerostat Radar System vessel and object detection United States — — Proposed
US Customs and Border ProtectionRAPTOR (Rapid Tactical Operations Reconnaissance) United StatesAlong US borders — — Proposed
Bureau of Alcohol, Tobacco, Firearms and ExplosivesAirship AI video management platform with licence plate OCR and object detection United StatesNationwide Airship AI Holdings Inc. — Operational
NASA Office of Protective ServicesAI camera analytics for security United StatesNationwide Milestone, Genetec, Verkada, Lenel 2016 Operational
US Department of Veterans Affairs (Office of Operations, Security and Preparedness)Avigilon video search function United States Avigilon (named in the use case title) — Operational
US Department of Veterans Affairs (Office of Operations, Security and Preparedness)Verkada camera AI video search United States Verkada (named in the use case title) — Proposed