I recently saw the title of a Recorded Future podcast regarding AI and police reporting. I have 28 years of law enforcement experience, 8 years as a uniformed police officer and this title really intrigued me. So I watched the segment: AI is Writing Police Reports, Should We be Worried?[1]
The story starts with police body cams, which began somewhat experimentally in 2011 and now has gain acceptance throughout US policing. The main purpose of demanding police wear body cams was to change police behavior, specifically in the use of force. With over 10+ years of police body cam usage, only a slight change in police behavior has taken place. I personally left uniform policing in 1991, when video was just beginning in police vehicles and also used in the arrest booking process. Back then I saw it as a positive, as it demanded higher levels of police professionalism.
The main provider of police body cams is Axon.[2] If you have ever watch actual police body cam footage, you most likely saw Axon on the bottom of the video footage. Recently, Axon developed an artificial intelligence (AI) program that takes body cam footage and develops a draft police report. This AI program is called “Draft One.” This new technology piqued the interest of Officer Matt Barter of the Manchester NH Police Department (MPD). When I heard Matt’s name, I thought “hey I know him.” I met Matt over 8 years ago when I was helping out at the New Hampshire Fusion Center. Matt was then handling the department’s crime heat maps and crime statistics for use in patrol operations. Matt reached out to Asst. Professor Ian Adams of University of South Carolina, who was himself a past police officer in Utah.
The Manchester Police agreed to conduct a study using Draft One for 6 weeks. The hypothesis was that utilizing Draft One would reduce time spent with officers writing reports. The reduction on hours of administrative type work could be a positive force multiplier for their department. MPD is a 250 officer sized police agency and is considered a medium sized police department in the US. The plan was to have half the department used Draft One and the other half with traditional reporting. After 6 weeks, the study showed no significant reduction of time spent on report writing. In this study, the technology did not prove to be a positive resource. Both tech and law enforcement professionals are now claiming that there is no current language learning model (LLM) that will write a police report. And regrading this AI technology, not all police departments in the US have body cams (which power the AI report writing tool).
Officer Bob Unger, with the Ft. Collins CO Police Department (FCPD) counters the Manchester study. Unger said that when he started with FCPD close to 23 years ago, officers would dictate police reports on a microcassette and deliver to a steno pool, who would write the reports for the officers. This I thought intriguing (I have to write reports long hand, or type them out with an old Royal typewriter), as the FBI handling report writing in this fashion for years. The turn around for report completion at FCPD was 7-10 days, not really a model of efficiency, but did get officers back on the streets versus sitting behind a typewriter. Because of this past practice, FCPD was interested in the Draft One technology.
Unger said that with their current use of Draft One, that officers really like the application. FCPD is only using the app for low level criminal reports, this to flush out any tech issues in the usage of this AI tool. Unger says that the app is saving manhours in report writing and getting officers back to active policing.
Professor Adams suggests that Draft One technology should be more focused on improving report writing versus the promotion time saving of report writing. Adams explains that police officers are human and at times hurry through report writing for a variety of human reasons. Many police sergeants (who traditionally are the first editors of police reporting) believe that the usage of AI technology for report writing will improve the quality of the report. And then the savings of time would/could be at the editing level of police operations. Police reports must be accurate and fair and AI technology could help with report quality.
The Problems: there is a theory that a technology company like Axon is a publicly held company that demands profit. Some question the goal of company profits, set against the actual quality of the product. In this case, a vital function of police operations; report writing. Silicon Valley has proven time and time again, that profits (quick profits) often compromise quality, and more specifically cyber security components. Axon started with providing body cams, which again did not specifically provide the initial anticipated results of changing police behaviors, but the body cams actually have helped justify many legal and ethical police actions. But to drive more profits at Axon, they developed Draft One. A product that is still being studied.
Then there is government oversight. Traditionally all levels of government are woefully behind placing regulations and statues on lightening speed technology. The US Congress has oversight on federal law enforcement and state governments with state, county and local police agencies. Police are no different than the general public in that gaining an advantage in operations via technology - is tempting. Sometimes the testing of these type police related products lack proper evaluation and sound oversight. Police - Buyer beware.
Personally, I am not convinced technology will replace human involvement, especially in police work. It will always be a human to human interaction. But tools that assist officer’s administrative workload should be explored, specifically in writing reports. An officer will not totally appreciate a good report, until they are called to testify in a criminal court and must rely on their 2+ year old police report.
This article is shared at no charge for educational and informational purposes only.
Red Sky Alliance is a Cyber Threat Analysis and Intelligence Service organization. We provide indicators of compromise information via a notification service (RedXray) or an analysis service (CTAC). For questions, comments or assistance, please contact the office directly at 1-844-492-7225, or feedback@redskyalliance.com
Weekly Cyber Intelligence Briefings:
- Reporting: https://www.redskyalliance.org/
- Website: https://www.redskyalliance.com/
- LinkedIn: https://www.linkedin.com/company/64265941
Weekly Cyber Intelligence Briefings:
REDSHORTS - Weekly Cyber Intelligence Briefings
https://register.gotowebinar.com/register/5378972949933166424
[1] https://podcasts.apple.com/us/podcast/170-ai-is-writing-police-reports-should-we-be-worried/id1225077306/
Comments