Deputies are now allowed to use Axon Draft One AI to generate some preliminary police reports, Hillsborough County Sheriff’s Office announced Thursday.
Critics of the software say that police contractor company Axon designed the tool to prevent anyone—including media, defense attorneys and civil liberties organizations—from telling exactly which parts of a report were composed by AI and which were written directly by the officer.
Draft One uses audio recordings of police interactions from a body camera to draft reports. Since video isn’t used by the AI, an officer approaching a red vehicle would have to state as much out loud and read out the license plate if they wanted to ensure Draft One captures it in the report.
Prior to the implementation of Axon Draft One a deputy responding to a non-criminal call for service would type out a report of what happened in their own words. Now, that deputy goes back to the vehicle and let AI draft a report, which will then be edited by the officer based on prompts provided by the system. Deputies must certify that reports match their own account of what happened.
The Anchorage Police Department discontinued use of Draft One in 2024 after finding it inefficient. “We were hoping that it would be providing significant time savings for our officers, but we did not find that to be the case,” APD Deputy Chief of Administration Gina Burington said at a public meeting.
Because every detail had to be verbally noted to be included in AI’s summary, Burlington said the editing process took about as long as writing a report from scratch.
For now only non-criminal reports are being drafted with AI. But police departments continue to expand use of Axon products like Draft One every year, driving quarter-over-quarter stock price increases for the company, including a 17.5% jump just yesterday. Axon’s sales reps are given high quotas and get smaller paychecks if they don’t hit them, according to postings on employer review sites like Glassdoor.
Digital civil liberties organization Electronic Frontier Foundation (EFF) compiled a comprehensive report on how difficult it is to audit the performance of Draft One. The findings are titled “Axon’s Draft One Is Designed to Defy Transparency.”
EFF requested Draft One records from several police departments, only some of which were actually able to compile a list of reports generated with Draft One. This is because not every agency requires a report to disclose if Draft One was used to generate it. A representative for HCSO told Creative Loafing Tampa Bay that she is looking into the office’s disclosure policy, and this article will be updated once we hear back.
EFF also found that when departments do disclose AI was used, it’s impossible to tell which parts of the report were generated by AI, and which edits (if any) were made by the deputy. This is because Axon intentionally does not retain the drafts generated by AI.
“We don’t store the original draft and that’s by design,” Axon’s senior principal product manager said in a roundtable discussion about the product. “That’s really because the last thing we want to do is create more disclosure headaches for our customers and our attorney’s offices.”
EFF’s report says the lack of draft retention makes it harder to assess the performance of the AI bot. It’s not possible to detect large-scale bias or errors if you can’t even compile an accurate database of information composed by Draft One, EFF says.
Former state prosecutor and current defense attorney Francesca de la Grana told CL that police should always be disclosing AI use and reviewing cases carefully.
“A police report is a direct account of the events that occurred. As a defense attorney, we rely directly on the police reports to identify what happened in a case. I don’t see how a police report could be generated by AI,” de la Grana added.
De la Grana noted that, per her preliminary impression, the way Draft One is being used is seemingly OK for now as long as the officer is actually editing the report.
By Axon’s design, however, it’s impossible to know the degree to which the officer has edited it.
De la Grana said that while AI can be a helpful tool for efficiency, it’s important that deputies take full responsibility for the accuracy of their reports rather than blaming the AI tool for any issues.
“You are a law enforcement officer, and you have to swear that this report is true and accurate,” de la Grana told CL. “If you didn’t review it and sign off on it, then it’s not.” She warned that if a department becomes overly reliant on AI to write reports, it could diminish an officer’s credibility in court.
California created a new law last year to require that the first draft of any AI-generated police reports be retained, contrary to Axon’s policy.
EFF says police should not use AI to write reports at all.
“There are just too many unanswered questions about how AI would translate the audio of situations and whether police will actually edit those drafts, while simultaneously, there is no way for the public to reliably discern what was written by a person and what was written by a computer,” EFF’s report reads. “This is before we even get to the question of how these reports might compound and exacerbate existing problems or create new ones in an already unfair and untransparent criminal justice system. … Anyone with control or influence over police departments, be they lawmakers or people in the criminal justice system, has a duty to be informed about the potential harms and challenges posed by AI-written police reports.”
Pitch in to help make the Tampa Bay Journalism Project a success.
Subscribe to Creative Loafing newsletters.
Follow us: Google News | NewsBreak | Reddit | Instagram | Facebook | BlueSky
This article appears in Feb. 26 – Mar. 04, 2026.