Officer Scott Brittingham of Fort Collins, Colorado, began using AI-powered software, Draft One, to expedite police report drafting, reducing his report writing time from 45 minutes to 10 minutes.
Officer Scott Brittingham, with nine years of service at the Fort Collins, Colorado, police department, initially expressed pride in his meticulous process of drafting reports following each call for service. Despite his established routine, he indicated intrigue when the department initiated a test of a new tool designed to enhance efficiency in this particular task. This new methodology has resulted in a significant reduction in the time required for report completion; a report that previously necessitated 45 minutes of his attention now requires only 10 minutes for full completion.
Brittingham articulated a degree of initial skepticism, stating in a March interview at the Fort Collins police station for CNN’s Terms of Service podcast, “I was a little bit skeptical, I’m not a big technology person.” He further elaborated on the perceived benefit, noting that dedicating less time to report writing enables him to “take more calls for service” and “be proactive in preventing crime.” This direct correlation between reduced administrative burden and increased field presence represents a key operational advantage highlighted by Brittingham.
The technology referenced by Brittingham is Draft One, an artificial intelligence-powered software designed to generate the initial draft of police reports. The primary objective of this integration is to accelerate and simplify the report writing process for law enforcement personnel. The adoption of such tools, exemplified by Brittingham’s experience, is projected to become an increasingly prevalent practice among police officers as departments nationally implement similar systems.
The increasing traction of Draft One occurs amidst ongoing discussions and concerns raised by segments of the legal community and civil rights organizations. These concerns primarily revolve around the potential for AI-drafted police reports to contain inherent biases or factual inaccuracies. Additionally, questions regarding transparency in the generation and modification of these automated reports have been presented. The implications of these concerns extend to the integrity and impartiality of the criminal justice system, necessitating rigorous evaluation of the technology’s deployment.
Axon, a prominent law enforcement technology company known for manufacturing tasers and body cameras, developed Draft One. The company has characterized Draft One as its most rapidly expanding product since its introduction last year. Axon is not the sole entity operating in this technological segment. Truleo, another law enforcement technology provider, offers a comparable AI-driven police report tool identified as Field Notes, indicating a developing competitive landscape within this specialized market.
Police reports constitute a foundational element within the criminal justice framework. Officers utilize these documents to provide detailed accounts of incidents and to articulate the rationale behind their actions taken during the course of duty. These reports may subsequently serve as preparatory material for officers required to provide testimony in court proceedings. Furthermore, these reports furnish critical information to prosecutors, defense attorneys, presiding judges, and the general public, delineating the officer’s perspective on the events that transpired. Andrew Guthrie Ferguson, a law professor at American University specializing in the intersection of technology and policing, emphasized that police reports can influence prosecutorial decisions regarding case pursuit or judicial determinations concerning pre-trial detention without bond.
Ferguson articulated that “Police reports are really an accountability mechanism.” He further stated, “It’s a justification for state power, for police power.” This perspective underscores the role of reports in documenting and legitimizing law enforcement actions, thereby serving as a critical component of institutional oversight and public trust.
Advocates of Draft One emphasize the potential for AI to enhance the accuracy and comprehensiveness of police reports, in addition to the acknowledged benefits related to time efficiency. Conversely, skeptics express apprehension that any inherent issues within the technology could precipitate substantial consequences for individuals involved in the legal system. At least one state has already enacted legislation specifically regulating the application of AI in the drafting of police reports, reflecting a proactive regulatory response to the emerging technology.
The introduction of Draft One coincides with broader concerns pertaining to the application of artificial intelligence within law enforcement contexts. Previous implementations of technologies such as facial recognition have resulted in documented instances of wrongful arrests, contributing to a cautious approach regarding new AI deployments. Ferguson acknowledged the expanding integration, stating, “I do think it’s a growing movement. Like lots of AI, people are looking at how do we update? How do we improve?” He also noted a commercial driver, adding, “There’s a hype level, too, that people are pushing this because there’s money to be made on the technology.”
The operational process for Draft One begins after an officer records an interaction using their body camera. The officer can then initiate a request for Draft One to generate a report. The software utilizes the transcript derived from the body camera footage as its basis for creating the initial draft. This draft typically becomes available for review within seconds of the request. Following the generation of the draft, the officer is prompted to review the content and incorporate any additional pertinent details before formally submitting the report as final.
Each draft report produced by Draft One incorporates bracketed fill-in-the-blanks. Officers are required to either complete these designated sections with specific information or delete them if they are not applicable. This structural element is intentionally designed to ensure that officers thoroughly review the automated drafts, thereby facilitating the correction of potential errors or the inclusion of any missing information. Josh Isner, President of Axon, affirmed the requirement for officer involvement, stating to CNN, “It really does have to be the officer’s own report at the end of the day, and they have to sign off as to what happened.”
Draft One operates on a modified version of OpenAI’s ChatGPT. Axon has further refined and trained this underlying AI model with the specific objective of mitigating the occurrence of “hallucinations,” which are defined as factual errors that AI systems can generate without a basis in the provided data. Axon also collaborates with an external group comprising third-party academics, restorative justice advocates, and community leaders. This collaborative effort aims to solicit feedback regarding the responsible development of its technology and to identify strategies for mitigating potential biases that could arise from its application.
The conceptualization of Draft One originated from the staffing challenges experienced by Axon’s police department clientele, according to Isner. A 2024 survey conducted by the International Association of Chiefs of Police, which encompassed over 1,000 US police agencies, revealed that these agencies were, on average, operating at least 10% below their authorized staffing levels. This data underscores a significant personnel deficit across the law enforcement sector.
Isner stated, “The biggest problem in public safety right now is hiring. You cannot hire enough police officers.” He further emphasized the strategic importance of efficiency tools, remarking, “Anything a police department can adopt to make them more efficient is kind of the name of the game right now.” This perspective positions technologies like Draft One as critical solutions for addressing widespread understaffing issues.
Axon declined to specify the exact number of departments currently utilizing Draft One. However, the software has been adopted by police departments in Lafayette, Indiana; Tampa, Florida; and Campbell, California. Isner indicated that “almost every single department” in the United States currently employs at least one Axon product, suggesting substantial potential for further market penetration for Draft One given the company’s existing widespread client base.
In Fort Collins, Technology Sergeant Bob Younger initiated a test of Draft One last summer after observing a demonstration of the tool’s capabilities. Younger recounted his immediate reaction, stating, “I was blown away at the quality of the report, the accuracy of the report and how fast it happened.” He further articulated his assessment, concluding, “I thought to myself, ‘This is an opportunity that we cannot let go.’”
The Fort Collins Police Department initially extended access to Draft One to approximately 70 officers. Currently, the technology is accessible to all officers within the department. Younger estimates that the implementation of the tool has reduced the time officers dedicate to report writing by nearly 70%. He quantified the benefit, stating, “and that’s time we can give back to our citizens,” highlighting the direct community impact of increased departmental efficiency.
Isner reported a largely positive reception from prosecutors regarding Draft One. However, in September of the previous year, the prosecutor’s office in King County, Washington, declared that it would not accept police reports generated with AI assistance. This decision followed expressions of interest from local law enforcement agencies in deploying Draft One. The King County office communicated its rationale to police chiefs via email, stating that utilizing the tool would “likely result in many of your officers approving Axon drafted narratives with unintentional errors in them.”
An Axon spokesperson affirmed the company’s commitment to ongoing collaboration, stating that Axon is “committed to continuous collaboration with police agencies, prosecutors, defense attorneys, community advocates, and other stakeholders to gather input and guide the responsible evolution of Draft One.” The spokesperson also specified that the AI model underpinning Draft One is “calibrated … to minimize speculation or embellishments,” indicating a design focus on factual accuracy.
Concerns regarding potential errors or biases in AI-drafted police reports are not limited to King County prosecutors. Jay Stanley, a policy analyst with the ACLU Speech Privacy and Technology Project, articulated his organization’s vigilance, stating, “When you see this brand new technology being inserted in some ways into the heart of the criminal justice system, which is already rife with injustice and bias and so forth, it’s definitely something that we sit bolt upright and take a close look at.” Stanley authored a report last year that recommended against the use of Draft One.
Even Ferguson, who anticipates the widespread adoption of this technology in policing, expressed apprehension regarding potential inaccuracies arising from body camera footage transcripts impacting reports. Ferguson specifically noted, “The transcript that you get, which becomes a police report, might be filled with misunderstandings, because the algorithm didn’t understand, like, a southern accent or a different kind of accent.” He further pointed out that nonverbal cues, such as a person nodding instead of verbally assenting, might not be accurately captured or reflected in the AI-generated report.
Axon has implemented a mechanism to mitigate errors or missing details through the inclusion of automatic blank fields within Draft One reports. However, during a demonstration at the Fort Collins Police Department, CNN observed that it is possible for an officer to delete all of these pre-defined prompts and subsequently submit the report without incorporating any modifications. Furthermore, once a report is formally submitted as final, the original, AI-generated draft is not retained, which means there is no audit trail to determine what specific changes, if any, an officer made to the initial AI-generated content.
Axon explained that the non-retention of original AI-generated drafts is intended to parallel traditional, manual report writing processes, wherein an officer’s handwritten drafts would not typically be preserved alongside the final report. The company also provides an opt-in setting that allows police departments to mandate a minimum percentage of edits to a draft before its submission. This feature enables departments to enforce a degree of officer review and modification.
A separate consideration involves the issue of transparency, specifically whether a defendant would be informed that the police report in their case was drafted with AI assistance. Final reports generated using Draft One include a customizable disclaimer by default, indicating that AI contributed to their creation. However, police departments retain the option to disable this feature. The Fort Collins Police Department, for instance, does not include such disclaimers. Despite this, Younger stated that officers are incentivized to personalize reports and ensure their accuracy.
Younger stated, “What an officer is worried about is being critiqued or held responsible for an error or doing something and being inaccurate.” He further emphasized, “Officers are super hyper-focused on the quality and quantity of their work,” suggesting that professional accountability drives their attention to detail regardless of the disclaimer’s presence. Ferguson, however, advocated for full disclosure, asserting that “radical transparency is the best practice.” In alignment with this principle, Utah state lawmakers enacted legislation earlier this year requiring police departments to include a disclaimer on all final reports that were drafted with AI assistance.
Officer Brittingham offered a concluding perspective on Draft One, characterizing it as a tool rather than a definitive solution. He stated, “My overall impression is that it’s a tool like anything else.” Brittingham further clarified its function, asserting, “It’s not the fix. It’s not replacing us writing reports. It’s just a tool to help us with writing reports.”
