Oklahoma City – The Oklahoma City Police Department has recently begun using AI technology to draft incident reports, a move that has generated both enthusiasm and concern within law enforcement and legal circles. The new tool, developed by Axon and named Draft One, leverages advanced machine learning algorithms similar to those used by ChatGPT to create detailed reports from body camera footage and audio recordings.
Sgt. Matt Gilmore, who recently utilized the tool during a search operation with his K-9 partner, Gunner, praised its efficiency and accuracy. "The report was better than I could have ever written, and it was 100% accurate," Gilmore remarked. The AI-generated report even included details he hadn't initially recalled, such as the color of a suspect’s vehicle mentioned by another officer.
Draft One, which processes sounds and radio chatter from body cameras, produces a first draft of a report in just eight seconds. This marks a significant time-saving advance for officers, who previously spent 30 to 45 minutes manually writing reports. Axon, renowned for its body cameras and Tasers, views this innovation as a "game changer" for police work, particularly in reducing the tedious data entry involved in report writing.
However, the introduction of AI-generated reports raises several concerns. Legal scholars, prosecutors, and police watchdogs worry about the potential implications for criminal justice. They fear that reliance on AI could compromise the integrity of police reports, which are crucial for legal proceedings and often form the basis for prosecutorial decisions.
Rick Smith, CEO of Axon, acknowledged these concerns. “While the response to Draft One has been overwhelmingly positive, there are legitimate concerns about ensuring that police officers, rather than AI, are responsible for their reports,” Smith said. He emphasized that district attorneys and legal professionals want to ensure officers are accountable for their testimonies, particularly in court settings.
The technology is currently limited to minor incident reports in Oklahoma City, excluding cases involving arrests, felonies, or violent crimes. Other cities, such as Lafayette, Indiana, and Fort Collins, Colorado, have adopted Draft One more broadly, though some challenges, like managing high noise levels, remain.
Beyond drafting text, Axon experimented with computer vision to summarize video footage but encountered limitations, particularly in addressing biases and ensuring sensitivity. This led to a focus solely on audio processing for the current version of Draft One.
As AI continues to integrate into policing, there are calls for a broader public discussion about its benefits and potential risks. Legal scholar Andrew Ferguson points out that the tendency of large language models to generate false information—known as "hallucination"—could introduce inaccuracies into police reports, raising questions about reliability and accountability.
For now, the adoption of AI in report generation is being monitored closely. Officers are required to acknowledge the use of AI in their reports, and as the technology evolves, it will likely continue to shape the future of police documentation.
