Law enforcement agencies: artificial intelligence.

Senator Arreguín, with Assembly Member Schultz as a coauthor, advances a measure that would require law enforcement agencies to adopt a formal policy governing the use of artificial intelligence in official reports, mandating per-page disclosure of AI involvement and a responsible officer’s signature on the final document, while preserving the first AI draft and creating an auditable record of AI usage. The proposal applies to any agency employing peace officers across state and local government, and it is described as a state-mandated local program with a reimbursement pathway if costs are found to be mandated.

Core provisions require that every official report identify the specific AI tools used on each page and include the statement that the report was written with artificial intelligence, alongside the officer’s or agency member’s signature verifying the contents. If AI is used to prepare an official report, the first AI draft must be retained for as long as the official report is retained, and AI-generated drafts are not themselves sworn statements. An audit trail must log the identity of the user who operated the AI and, where available, the video and audio inputs used to generate the narrative. Contracted vendors may not share or use agency data beyond the agency’s purposes or court orders, but may access data for troubleshooting, bias mitigation, accuracy improvement, or system refinement. Definitions set boundaries for terms such as “artificial intelligence,” “first draft,” “official report,” “contracted vendor,” and “law enforcement agency.”

Implementation hinges on local agencies adopting policies that incorporate these elements, updating standard operating procedures and training, and establishing IT capabilities to capture disclosures, signatures, and audit trails; retention practices for first drafts must align with the duration of the official report’s retention. The measure contemplates potential costs to local agencies, with reimbursement governed by the state’s mandate framework if the Commission on State Mandates determines costs are mandated by the state, though there is no explicit funding within the bill. Enforcement provisions are not specified in the text, with oversight anticipated to occur through existing mandate reimbursement processes and internal compliance mechanisms.

Contextual considerations include how AI-disclosed, signed official reports and the accompanying audit trails may affect evidentiary authentication, discovery, and public-records practices, as well as the data governance implications for contracted vendors. Stakeholders include law enforcement agencies, district attorneys, courts, and technology providers, all of whom would navigate new policy requirements, retention obligations, and contract modifications. The bill does not specify an effective date or phase-in schedule, leaving implementation details to standard statutory practice. Overall, the measure codifies governance around AI-assisted report-writing and data handling in policing, emphasizing disclosure, traceability, and vendor controls.