Honolulu police are about to find out what happens when artificial intelligence starts listening in on bodycams. For 30 days, the Honolulu Police Department will flip on an automatic transcription feature that funnels officers’ body-worn camera audio into software that spits out draft incident narratives. HPD says officers will not be able to log into the draft-writing platform during this short trial and that human reviewers will still sign off on every final report. Civil liberty advocates and union leaders are already warning that a narrow test like this can be the first step toward something much bigger in the surveillance department.
What HPD Is Testing
An internal notice obtained by Civil Beat says the auto-transcription feature will be turned on for 30 days so the department can "evaluate the accuracy and potential time savings" of using AI to help write police reports. The technology on the table is Axon’s Draft One, which takes body-worn audio and generates draft narratives. In a press release, the Honolulu Police Department said it is "beginning to explore" AI tools and plans to invite public input before anything gets rolled out more broadly. HPD also stressed that any pilot would still require officer review and would not replace officer judgment, and earlier debates about the idea were covered in earlier discussions.
Why Advocates Are Worried
Lt. Joseph O'Neal of HPD's information-technology division described the system to Civil Beat as "just fancy dictation." In his telling, the tool simply transcribes body-cam audio and turns it into narrative summaries that officers must still edit and approve. Union leaders and community advocates are not so reassured. Nicholas Schlapak, president of the state police union, warned that the same tool could be repurposed to monitor officers, predict behavior or even "compromise our safety." Community organizer Liam Chinn of the Reimagining Public Safety in Hawaiʻi Coalition called the trial a potential gateway to "surveillance creep." Their concerns focus on how accurate these AI drafts will be, how long the drafts are kept and how any automated text might be used later in investigations and prosecutions.
How Other Agencies Have Responded
Honolulu is hardly the first department to experiment with AI-generated police reports, and the early reviews elsewhere are mixed. Anchorage police ran a 90-day trial and then decided not to move forward, after finding that the time officers spent editing AI-generated drafts wiped out the supposed efficiencies, according to Alaska Public Media. On the other hand, pilots in places such as Fort Collins and other jurisdictions have reported significant time savings. Reporting and federal guidance also note that these systems lean entirely on audio transcripts and cannot interpret visual cues in video. That limitation raises obvious questions about accuracy and how the resulting narratives will hold up as evidence, issues explored in coverage from KQED and background from the U.S. Department of Justice's COPS office.
Legal And Policy Questions
Prosecutors and lawmakers elsewhere are already drawing some bright lines as AI-generated narratives creep into policing. The King County prosecuting attorney's office has said it will not accept police narratives that are produced with AI at all, as reported by WPBF. California has gone a different route, passing S.B. 524 to require agencies to disclose when AI was used in drafting reports and to preserve the very first AI draft for later auditing. Advocacy groups such as the EFF and policy organizations like EPIC have pushed for strong transparency rules and strict retention requirements as these tools spread.
What's Next For Honolulu
For now, Honolulu’s experiment is limited to that 30-day test, but HPD already has its eye on what comes after. The department says it has notified the Prosecuting Attorney's Office and will seek community input while it measures how accurate the system is and how it affects officers’ workloads. According to HPD, a fuller pilot could start as early as Spring 2026. The department emphasizes that any AI output would still need human verification and says it is evaluating multiple vendors and possible safeguards. On the other side, advocates and union leaders say they plan to push hard for audit trails, public oversight and firm limits on how body-camera data and AI-generated text can be stored, shared or used in the future.
