Technology’s Effect on Advocacy
Intelligence (AI). ”) If your cases involve LASD or any other large California agency, this is not a hypothetical. It is here.
that draft must be preserved. And it is discoverable.
Fourth, the agency must maintain an audit trail that identifies who used AI to create the report and what video or audio footage was used in its generation. (Pen. Code, § 13663, subd. (c).) Fifth, AI vendors are prohibited from selling, sharing, or using data submitted by law enforcement agencies for any purpose other than the agency’s needs or a court order, with a limited exception for troubleshooting, bias mitigation, accuracy improvement, or system refinement. (Pen. Code, § 13663, subd. (d).) California is only the second state, after Utah, to impose specific transparency requirements on AI- written law enforcement documents.
California’s Response: Senate Bill 524.
California has taken the lead on regulating this technology, and every practitioner needs to know what the law now requires. Senate Bill 524, authored by Senator Jesse Arreguín and signed by Governor Newsom on October 10, 2025, added section 13663 to the Penal Code. Effective January 1, 2026, the statute imposes five core requirements on any law enforcement agency that uses AI to generate official reports. First, on each page of the official report, or within the body of the text, the agency must identify every specific artificial intelligence program used in a manner that makes such identification readily apparent to the reader and must prominently state: “This report was written either fully or in part using artificial intelligence.” (Pen. Code, § 13663, subd. (a)(1).) Second, the report must bear the signature of the law enforcement officer or member who prepared it, in physical or electronic form, verifying that they reviewed its contents and that the facts contained in the report are true and correct. (Pen. Code, § 13663, subd. (a)(2).) Third — and this is perhaps the most consequential provision for litigators — the agency must retain the first draft created by the AI for as long as the official report is retained, and a draft shall not constitute an officer’s statement. (Pen. Code, § 13663, subd. (b)(1), (b)(2).) Before Senate Bill 524, AI vendors were designing their products to delete the original draft the moment the officer closed the browser window. Now
The Transparency Problem.
Senate Bill 524 was enacted for a reason. Before the law took effect, there was no meaningful way to audit AI-generated police reports. And here is the part that should alarm every civil rights practitioner: the dominant vendor, Axon, designed it that way. In July 2025, the Electronic Frontier Foundation (EFF) published an investigation based on public records obtained from police agencies already using Draft One. What they found was alarming. When an officer generates a report through Draft One, the AI produces a draft. The officer can edit it, add to it, or delete portions of it. But when the officer copies the final text into the department’s records management system and closes the Draft One window, the original AI-generated draft disappears. No version history. No record of which portions were written by AI and which were written by the officer. Gone.
29
Consumer Attorneys of California
FORUM September/October 2026
Made with FlippingBook - Online catalogs