AI Frontier Institute

Requires policing agencies to conduct an inventory of, and develop a publicly-available policy for, any artificial intelligence used to aid criminal investigations.

View primary source →

New York Assembly Bill 9253-A would add a new Section 837-ww to the Executive Law, requiring policing agencies to disclose and develop policies around any artificial intelligence used to aid criminal investigations. 'Covered AI' is broadly defined to include technologies such as biometric identification, facial recognition, license plate readers, predictive policing, gunshot detection, social media analysis, and generative AI used to write police reports, but excludes purely administrative tools like spell-checkers. Under subdivision 2, agencies must publish a publicly available inventory of each covered AI system at least 90 days before deploying a new system, or within 180 days for systems already in use, including vendor names, capabilities, limitations, data inputs and outputs, authorized uses, and internal validation summaries. Under subdivision 3, any use of covered AI in a criminal investigation must be disclosed in the police report, and shared with prosecutors and defense counsel. Reports generated using generative AI must include an AI disclaimer, author certification of accuracy, retention of first drafts, and an audit trail. Under subdivision 4, the Attorney General must publish a model policy within 90 days of enactment and review it annually; each police agency must then adopt that policy or an equivalent within 90 days of its issuance and submit it to the Attorney General. Under subdivision 5, the Attorney General may bring civil enforcement actions, and residents may also sue for equitable or declaratory relief to enforce the inventory and policy requirements, with prevailing plaintiffs entitled to attorney's fees, provided 90 days' prior written notice is given to the agency.

← Back to the tracker