Technology is widely used in police work today, with license plate readers, facial recognition software, and other tools that enhance law enforcement efficiency and effectiveness.

Now some agencies are using Artificial Intelligence (AI) to optimize officers' time in the field by shortening the time and improving the quality of police incident reporting with software that connects to body cameras, uploads the footage, and transcribes the audio to generate fully detailed reports in minutes

The AI generative software that supports this technology is currently available from two companies, AXON and Truleo. AXON’s voice transcription software, Draft One, is based on ChatGPT, made by San Francisco-based OpenAI. Truleo’s report writing software, Field Notes, is powered by Amazon Bedrock, a product of Amazon Web Services (AWS).

A Time-Saving Force Multiplier

Many of the growing number of law enforcement agencies who use AI transcription describe it as a game changer, saying it saves hours of time normally spent on report writing and frees officers to focus more on policing. These users see it as a force multiplier.

They also say that AI transcription ensures that police reports are clearly written and capture all of the verbal interchange surrounding an incident, even speech that the officer did not notice or does not remember hearing.

Said Jason Lucas, a detective with the Oklahoma City Police Department (OKCPD), “This technology can not only save time, but keep us more accountable. It has also helped me improve my own writing skill, even when I am not using Draft One.”

Chief Stephen Redfearn of the Boulder (Colorado) Police Department (BPD) is equally enthusiastic. “AI–written reports is one of the most exciting changes in policing technology that I have seen,” he said. “It’s substantially reduced the time our officers spend writing reports, freeing them up for more proactive policing. I think it improves overall report quality too.”

Legal and Community Concerns

But despite the rave reviews, there are community concerns about transparency and bias in machine learning, as well as questions about the admissibility of AI written reports in court proceedings; some doubt that prosecutors and defense attorneys will accept an AI report’s findings as evidence.

According to OKCPD Captain Jason Bussert, the department had a meeting with the District Attorney for each of the four counties they cover as well as a representative from the U.S. Attorney’s office for the Western District of Oklahoma. As a group, they advised caution before using it on high-stakes criminal cases.

“Half of them liked [Draft One] and the others were hesitant because they wanted to see how defense attorneys attacked the report and how it played in court,” he said. So, for now, they only use it for minor incidents that don't lead to arrests.

But other law enforcement leaders note that a poorly written report can weaken the prosecution’s case, while a well-written report produced by AI can result in more accountability and justice for victims. “It makes us look more professional, too,” added Chief Redfearn.

As for concerns about bias influencing an AI–generated narrative, Redfearn said, “The video can be viewed to support what is transcribed. If for instance, the officer says, “he appears to be drunk,” the camera footage will either validate or undercut that.”

In the words of Detective Lucas, “Everything the AI–generated report says was heard; it’s not creating evidence.”

To address the concerns that important questions or critical details might be overlooked, both Truleo and AXON software provide templates for each type of incident with related information prompts in the generated report.

Said Truleo CEO, Anthony Tassone, “If it’s a domestic call for instance, Field Notes will ask the officer to document if there are children, their names, birth dates, and any injuries. There is also a requirement that officers review their report, manually fill in missing information, and then sign off on the narrative’s accuracy before submission”.

To address these and other concerns, BPD Chief Redfearn met with Boulder’s Police Oversight Panel, which reviews investigations into complaints of officer misconduct and advises on department policies. He also met separately with community members to address concerns about bias.

Since launching the program in January 2024, Redfearn has seen no problems with this or other concerns. “We started out using Draft One for minor incidents,” he said. “But after talking to District Attorneys, who said they were impressed with the quality of the reports, we decided to use it for everything.

“Officers used it during a violent domestic case as well as a sex assault case. And the detective used these reports for follow-ups and interviews. It’s accurate in recollection and can be used in any number of cases.”

Safeguards and Training

“We piloted the program for a year and are now rolling it out to all sworn personnel,” said Chief Redfearn “We’ll do monthly audits of the reports and compare them to body camera footage to ensure that they’re using it properly. A first line supervisor will look over the officer’s report to make sure it conforms to what is seen. As with any report, the officer has to sign off on it. But we want the community to know somebody is looking at this on a regular basis.”

Asked about AI’s ability to understand jargon, foreign language, or heavily accented English, Chief Redfearn said “AXON showed us a video of a cop with a very heavy southern accent and the AI–transcribed report got everything he said correctly. But there can be a problem with non-English speakers.”

According to Captain Bussert, a report generated by AI for an OKPD officer said that the victim had interactions with somebody who spoke no English. The report had no dialogue for the non-English speaker, but did capture the words of the victim, who did speak English. And Bussert thinks that AXON may be working on making Draft One capable of translating different languages.

Yet some police may not feel comfortable using AI to write their reports. In response to their concerns AXON states that, unlike consumer tools, Draft One is a public safety–grade AI solution designed specifically for police report writing and calibrated to stick to the facts within the narrative, without any embellishment.

Similarly, Amazon Bedrock, which supports Truleo’s Field Notes, states that it has "Guardrails" to help detect and mitigate errors through checks on generated narratives.

But training is critical. Detective Lucas said “It’s a great tool if you learn how to use it properly. I think it benefits any department to have an in-service training where somebody acts out a scene and you hand-write a report, then let the bodycam write out a report and see the difference in time and quality.”

Which is what OKCPD did before they went live with Draft One. “We had our officers compare their report with the AI one. About half the time, no changes had to be made; the other times, they had to fill in some missing information. The cameras show details that the officer may not notice or feel was important at the time, so new users need to learn to fully narrate the incident, leaving nothing out,” he said.

Funding for AI

A common concern is funding for new technology. As for Draft One, Chief Redfearn said “Its cost is charged per user. But I think it’s well worth it.

“When I went to the city to request funding, one of the things that impressed them was the reduction in time that officers have to sit and write reports. And if they’re busy, they might have to wait to the end of their shift and work overtime to complete them.

“This reduces the need for overtime pay and increases the amount of time officers are available to focus on crime and quality of life. When your staffing numbers are down, AI reporting helps a lot. This is the best way to present it to your funding source.”

Faye C. Elkins
Sr. Technical Writer
COPS Office

Tweet