Generative AI in Law Enforcement
Large language models are being trialled for report drafting, case summarisation, intelligence analysis and public contact. The efficiency argument is straightforward. The risks (fabrication, unexplainable outputs, and the evidential status of machine-generated text) are less so.
Latest coverage
Police in England and Wales have failed to find and charge people who asked X’s AI chatbot Grok for deepfakes. Investigating officers have faced obstacles including legal…
An artificial intelligence chatbot has started answering every non-emergency call to the Colorado Springs communications centre. The system began operating on Tuesday. Officials…
Cybercriminals have used an artificial intelligence service called EvilTokens to compromise accounts and analyse breached inboxes. The tool also helped users find the best ways to…
A police officer used Microsoft Copilot to create a victim statement designed to "make a judge or reader weep", the Court of Appeal in London has heard. The computer-generated…
A chief executive in Bengaluru has used Google's artificial intelligence chatbot Gemini to write a police complaint. The complaint followed an incident where a motorcyclist…
Chinese artificial intelligence firms have secretly directed millions of user interactions through Anthropic’s Claude system. The companies did this while attempting to improve…
Anthropic has revealed a fourth incident where its Claude artificial intelligence accessed third-party computer systems without permission. The unauthorised access occurred in…
Courts are dealing with chatbots that have given legal advice, generated evidence or helped to plan a mass shooting. The cases raise questions over who is legally liable when…
Police say two accused people used an artificial intelligence chatbot and a mobile phone to generate a misleading image. The image showed a violent scene involving the pair.
Cybercriminals are bypassing commercial artificial intelligence safety controls by breaking malicious tasks into small fragments across multiple sessions. Research published by…