If you were suspected of a crime, would you trust a chatbot to accurately convey the details of the incident? Some police departments believe the technology is ready for such tasks. In June, the Frederick Police Department in Colorado claimed to be the first law enforcement agency to fully implement Axon Draft One, an AI tool designed to produce police reports almost immediately after a body camera records an interaction.
Powered by OpenAI’s GPT-4 model, Axon Draft One summarizes the audio captured by body cameras. While this may seem practical, legal and civil rights experts caution about the potentially grave consequences. They warn that police reports serve as foundational documents in the justice system, influencing plea deals, sentencing, and more. With AI-generated reports, there are concerns about accuracy and accountability.
Andrew Ferguson, a legal expert, warns in his law review article that these chatbots may hallucinate or insert incorrect information, thus undermining the credibility of police reports. Despite these concerns, Axon claims that the AI tool is programmed to avoid such errors by limiting its ‘creativity’ settings.
Matthew Guariglia, a senior policy analyst at the Electronic Frontier Foundation, wrote a blog post urging for caution and rigorous scrutiny as more departments adopt this technology. While Axon emphasizes the efficiency of Draft One in reducing officers’ paperwork, experts continue to debate the ethical implications.
More details can be found at Ars Technica.