Police Test AI-Generated Reports as Legal Experts Raise Concerns

Compiled by the editorial desk with reference to official statements, public filings, and industry data.

Police departments in Colorado, Indiana, and Oklahoma are piloting an AI tool that drafts official reports from body camera audio, a move that has sparked warnings from legal experts about the reliability of machine-generated documents that can influence criminal cases.

The tool, called Draft One, was introduced in April by Axon, the company known for manufacturing tasers and other law enforcement equipment. Axon says the software, powered by OpenAI's GPT-4, can turn an officer's spoken narrative into a formatted report, potentially halving the time spent on paperwork. CEO Rick Smith told Forbes that if an officer spends half their day on reporting, cutting that in half could free up 25 percent of their time for patrol duties.

However, the adoption of AI in this context raises questions because police reports are not routine correspondence; they serve as foundational documents in investigations and legal proceedings. Generative AI systems like GPT-4 are known to occasionally produce "hallucinations"—fabricated or incorrect details—which could have serious consequences if embedded in official records.

Andrew Ferguson, a law professor at American University who authored the first law review article on AI-generated police reports, told the Associated Press that he worries automation might make officers less meticulous in their writing. "I am concerned that automation and the ease of the technology would cause police officers to be sort of less careful with their writing," he said.

Axon defends Draft One's design, with AI product manager Noah Spitzer-Williams noting that the company has more control over the model than typical ChatGPT users. He said Axon has disabled GPT-4's "creativity dial" to limit the tool's tendency to embellish or invent details.

Broader Implications for Justice

Despite these safeguards, the integration of AI into police work is not without precedent for harm. Previous deployments of unproven AI systems in law enforcement have already led to tragic outcomes, underscoring the stakes of relying on automated tools in sensitive contexts.

Ferguson's review, published last month, poses a critical question: "The open question is how reliance on AI-generative suspicion will distort the foundation of a legal system dependent on the humble police report."

As more departments experiment with Draft One, some are even allowing its use for all case types, not just minor incidents. This expansion has amplified calls for public debate about what is lost when human judgment is delegated to algorithms.