Your AI Police Report Writing Policy: What It Needs to Say Before the First AI Assisted Report Reaches Your District Attorney
Updated: 9 minutes ago
An AI police report writing policy needs to answer six questions before the first AI-assisted report leaves your building: what the disclosure says and where it goes, whether the first draft is kept, who the audit trail names, who reviews the report, who signs and certifies it, and what your district attorney has agreed to accept. Two states already require most of that in writing. The rest of the country does not, yet, and it does not matter much, because the first person outside your department to read that report works for the prosecutor. If the policy is not written by then, the conversation happens on their terms instead of yours.
What an AI Police Report Writing Policy Has to Cover
Six clauses. Write them in this order, keep them short, and keep product names out of them. A policy with these six clauses addresses the core requirements in the two states that have acted, and it gives you a defensible foundation in front of a prosecutor, a city attorney, and a defense attorney in a state that has not.
Disclosure: Every report drafted fully or partly with artificial intelligence identifies the program used and carries a plain statement that artificial intelligence was used.
Retention of the first draft: The draft the system produced is kept for as long as the final report is kept.
Audit trail: The record identifies who used artificial intelligence to create the report, and it stays tied to the report.
Supervisor review: Who reviews an AI-assisted report, what they are reading for, and what they do when something is wrong.
Officer signature and certification: The officer who prepared the report signs it and certifies that it is accurate.
A written position agreed with your prosecutor: What your district attorney will accept, and what they want disclosed and preserved.
What the Law Already Requires in California and Utah
Two states have enacted requirements, and both are in force. Utah SB 180 became law on March 25, 2025, and took effect May 7, 2025. California SB 524 was signed October 10, 2025, and took effect January 1, 2026. There is no federal standard, so everything outside those two states is a patchwork, and more legislatures are drafting.
California is the more detailed of the two. Penal Code section 13663 requires each law enforcement agency to maintain a policy covering artificial intelligence use in official reports. Where artificial intelligence drafts a report fully or in part, every page of the report, or the body of the text, must identify every specific artificial intelligence program used in a way that is readily apparent to the reader, and prominently state: "This report was written either fully or in part using artificial intelligence." The statute also requires the signature of the officer or agency member who prepared the report, an audit trail that identifies at minimum the person who used artificial intelligence to create the report and the video and audio footage used to create it, if any, and retention of the first draft for as long as the official report is retained.
Utah SB 180 works at the policy level. Each agency has to have a policy on generative AI use that specifies which technologies are permitted and for which tasks. Any police report or law enforcement record created wholly or partially with generative artificial intelligence carries a disclaimer, and the author has to certify that they read and reviewed it for accuracy.
Read both as a floor, not a ceiling. Neither statute tells you how to run supervisor review, and neither one talks to your prosecutor for you.
The Clause Most Policies Miss: Keeping the First Draft
The draft the system produced and the report your officer filed are two different documents, and the difference between them is the record of what the officer actually reviewed and changed. California now requires the first draft to be kept for as long as the final report. Keep it everywhere else too, because that gap is the first thing a defense attorney will ask about.
This is the clause that gets missed, and the reason is boring. On many platforms, whether drafts are retained is a configuration setting, and settings get left at their defaults. A default setting is not a policy decision. Write the retention rule into the policy, in words, then confirm the system is configured to match it and note who checks that configuration. If your digital evidence management system holds the report drafts, the retention schedule for those drafts belongs in the same policy as the retention schedule for the video.
The Audit Trail, and Who Your Report Says Wrote It
Your audit trail is the transparency record, and it has to answer one question cleanly: who used artificial intelligence to create this report. Tie it to the report, retain it with the report, and assume it will be produced in discovery and in a public records request.
In practice that means the AI-generated draft is tracked separately from the officer's final submission, so the agency can show what the system produced and what the officer changed. That is the standard LensLock has already published in its own guidance on AI police report writing, and it is the version of an audit trail that survives contact with a courtroom. A log that only says "artificial intelligence was used somewhere in this agency" is not an audit trail. A record that names the user, the report, and the draft is.
Supervisor Review, the Officer's Signature, and Who Is Accountable
The officer who prepared the report signs it and certifies its accuracy. Supervisor review sits on top of that as a second check, and it does not move the accountability. Write both layers into the policy and say plainly which one carries the certification, because the two get blurred fast once reports start moving.
Real adopted policy already does this. Lafayette Police Department Policy 608 has officers sign to indicate they are aware transcription-assisted report generation was used, then routes the report for supervisor approval. Two steps, both recorded.
Be honest in the policy about the hard part. A supervisor reading for completeness is doing a different job than a supervisor reading for the narrative's accuracy. An inserted detail that reads perfectly well is exactly the kind of error that clears a review, so tell your supervisors what they are checking against, which is the body camera footage and the officer's own knowledge of the call, not the polish of the narrative. Critical details are the ones a generated narrative can hide best.
The Conversation to Have with Your Prosecutor Before the First Report Lands
Ask before you deploy. In a 2024 memo to its police agencies, the King County, Washington Prosecuting Attorney's Office said it would not accept police report narratives produced with the assistance of artificial intelligence, and that reports must be produced entirely by the authoring officer. The memo cited an AI-assisted report that was otherwise excellent but referenced an officer who was not at the scene. Your district attorney may have no position at all yet. That is the best possible time to shape one together.
Four questions to bring to that meeting:
Will you accept AI-assisted report narratives at all, and are there case types where you will not?
What disclosure do you want on the report itself, and in what wording?
Do you want the first draft preserved and produced in discovery?
How do you want artificial intelligence use handled when police officers testify in court?
Get the answers in writing, put them in the policy, and give the prosecutor's office a copy of the policy. A filing deputy who already knows what your disclosure line means is not reading it for the first time when the case lands.
What to Write in Your AI Police Report Writing Policy If Your State Has No Law Yet
Write the same six clauses. Two states have acted, others are drafting, and your city attorney, your prosecutor, and the first defense attorney to cross-examine your officer are not waiting for your legislature to catch up.
This is already normal practice, and it is public. Ogden Police Department Policy 75, Northfield Police Department Policy 344, Belchertown Directive 4.48, and Lafayette Policy 608 are all posted AI policies you can read before you write yours. Some agencies have gone further in the other direction: San Jose Police Department Order 2026-015 prohibits personnel from using artificial intelligence to generate official reports until the department has systems capable of meeting Penal Code 13663. That is also a legitimate policy position, and writing it down is the point either way.
Assume the policy gets read outside the department too. The Pasadena Community Police Oversight Commission has run a public survey on a proposed AI-assisted report writing policy, so community bodies are now part of the adoption conversation. For neutral third-party language you can borrow, the Policing Project at New York University School of Law publishes a model statute and policy guidance, and the United States Department of Justice COPS Office published a January 2025 Dispatch article with practitioner perspectives on AI-written reports.
Where Your Vendor Fits, and the Questions to Ask Them
The policy is yours to write. Letting officers use AI to write reports and writing the policy that governs it are one project, because most of the clauses depend on what your system can actually do. A public safety vendor that hands you a product and leaves the policy entirely to you has answered a question you did not ask.
Three questions to put to any vendor:
Can the system retain the first draft for as long as the final report, and is that a setting or a guarantee?
Can it produce an audit record of who used artificial intelligence on which report?
Can it carry the required disclosure onto the document itself, naming the AI transcription or drafting program used?
Where LensLock stands: AI report writing is live and generally available, an officer reads, reviews, and approves every report, and every draft is tracked separately from the final submission. The police body cameras and the evidence system are the same platform, so the footage, the draft, and the final report sit in one chain of custody instead of three.
Frequently Asked Questions
Does an officer have to disclose that AI helped write a police report?
In California and Utah, yes, by statute. Everywhere else it is not required by state law yet, but agencies are adopting disclosure as policy anyway. A prosecutor or a defense attorney is going to ask, and answering that question with a written policy is better than answering it on the stand.
What has to appear on an AI-assisted report in California?
Every page of the report, or the body of the text, must identify every specific artificial intelligence program used and prominently state: "This report was written either fully or in part using artificial intelligence." The preparing officer also signs the report certifying its accuracy. Both requirements are in Penal Code section 13663.
Do we have to keep the AI-generated first draft?
Under California law, yes. The first draft has to be retained for as long as the final report is retained. Keep it in any state, because the difference between the draft and the filed report is the only record of what your officer actually changed.
Will our district attorney accept an AI-assisted report?
It depends on the office, and at least one large prosecuting attorney's office has already said no. Ask before deployment, not after the first filing. Bring the four questions above, and put the answers in your policy.
What has to be in the audit trail for an AI-written report?
At minimum, the record has to identify the person who used artificial intelligence to create the report and the video and audio footage used, if any, and it has to stay tied to the report and be retained with it. Assume it will be produced in discovery and in public records requests, and build it so the draft is traceable next to the final submission.
Do we need a policy if our state has not passed a law?
Yes. Two states have acted, more are drafting, and your prosecutor, your city attorney, and your oversight body are not waiting for your legislature. Publicly posted agency policies already exist, so you are not writing from a blank page.
Who signs an AI-assisted report, the officer or the supervisor?
The officer who prepared it signs and certifies it. Supervisor review is a second layer on top of that certification, not a replacement for it. Say which is which in the policy so nobody has to guess when a report is questioned.
Before the First Report Goes Out
Write the six clauses, check your system against them, and go see your prosecutor. That order works. Running it backwards, product first and policy later, leaves the hardest questions until after the first report has already gone out.
Questions about how AI report writing, audit records, and draft retention work on LensLock? Call 866-536-7562, Press 2 to reach our Sales team. Press 3 to reach our Client Services team.
Comments