Requires policing agencies to conduct an inventory of, and develop a publicly-available policy for, any artificial intelligence used to aid criminal investigations.
This bill would add a new section to the Executive Law regulating the use of artificial intelligence by policing agencies in criminal investigations. It defines “covered AI” broadly to include tools that generate investigative leads or materially assist in police reports, such as biometric identification, facial or cross-camera tracking, automated license plate readers, predictive policing, risk scoring, gunshot detection, forensic DNA phenotyping, investigative genetic genealogy, social media analysis, and generative AI used in report writing. It excludes ordinary administrative tools like document management, spelling, or grammar checkers.
The bill requires any law enforcement agency using covered AI to conduct an annual inventory of all such systems and publicly post specified information about each system, including the vendor and product name, capabilities and limitations, data inputs and outputs, and authorized and unauthorized uses. Agencies must also disclose use of covered AI in police reports and, in certain circumstances, provide those reports to prosecutors and the person under investigation. If generative AI is used to create or materially assist a report, the report must include an AI disclaimer, identify AI-generated content where feasible, and include a certification that the author reviewed the report for accuracy; draft versions and audit trails must also be retained.
The bill further directs the Attorney General to issue a model policy within 90 days of enactment, review it annually, and require each law enforcement agency to adopt a policy meeting or exceeding that model. The model policy must limit AI use to expressly authorized purposes and state that violations may lead to administrative discipline. The Attorney General is authorized to enforce the law through civil action, and residents may sue for injunctive or declaratory relief to compel compliance with the inventory and policy-posting requirements, with attorney’s fees available to prevailing plaintiffs after notice and an opportunity to cure.
The bill’s impact would be to create statewide transparency, disclosure, retention, and policy requirements for police use of AI, while also giving the Attorney General and private residents enforcement tools. It would affect police departments, sheriffs, and other law enforcement agencies that use AI in investigations, as well as vendors whose systems are included in agency inventories. It would also likely shape how investigative records are drafted, retained, and shared with prosecutors and defendants.
Because there are no recorded committee transcripts or votes in the provided materials, there is no direct evidence of legislative debate or formal support/opposition. Based on the bill text, the measure appears aimed at accountability and oversight rather than expanding police powers, so its likely policy appeal is to transparency and civil liberties advocates, while potential concerns would center on administrative burden, disclosure of investigative methods, and operational constraints for law enforcement agencies.
The bill would amend the Executive Law by adding a new section governing law enforcement use of artificial intelligence in criminal investigations. It would impose annual inventory and public disclosure requirements, mandate AI-related disclosures in police reports, require retention and audit-trail practices for generative AI-assisted records, and compel adoption of an Attorney General model policy by policing agencies. It also creates enforcement mechanisms through Attorney General civil actions and private suits for certain violations, thereby adding new compliance obligations and potential litigation exposure for police agencies and political subdivisions.
No committee transcript or vote data were provided, so there is no recorded legislative sentiment to summarize from debate or roll calls. From the bill’s structure and caption, the measure appears to be framed as a transparency and accountability bill for police AI use, suggesting likely support from reform-oriented and civil liberties interests. At the same time, the absence of recorded discussion means there is no evidence here of formal opposition or negotiated amendments.
The main points of contention likely involve the breadth of the definition of “covered AI,” especially the inclusion of tools such as predictive policing, facial recognition, automated license plate readers, social media analysis, and generative AI for report writing. Law enforcement agencies may object to the administrative burden of annual inventories, public posting, disclosure deadlines, retention requirements, and audit trails, while privacy and accountability advocates may argue the bill does not go far enough because it regulates use and disclosure rather than banning certain high-risk technologies. Another likely issue is the private right of action and attorney’s fees provision, which could be viewed as a strong enforcement tool by supporters and as a litigation risk by opponents.