Skip to content
Submit incident
AII-1002-43669
Well documentedThree or more independent publishers report the core facts.

US police departments testing AI-generated incident reports

US police departments in Colorado, Indiana, and Oklahoma are piloting Axon's Draft One, an AI system that generates incident reports from body camera audio using GPT-4. Experts including legal scholars have raised concerns about reliability and the role of AI-generated reports in criminal cases, while some officers report the tool produces accurate results efficiently.

April 2024 ยท Oklahoma City Police Department ยท Reliability and hallucination
April 2024
AII-1002-436693 sources โ†“
Quick facts
Category
Reliability and hallucination
Country
United States
Source
The Independent
Date
April 2024
Drafted by
Raidu Incident Detection

What happened

Axon announced Draft One in April 2024, an AI tool built on OpenAI's GPT-4 that generates police incident reports from officers' body camera audio. Police departments in Colorado, Indiana, and Oklahoma have begun testing the system, with some allowing officers to use it for any type of case rather than only minor reports.

The tool is marketed as a productivity solution that can reduce officers' paperwork hours. An Oklahoma City police sergeant reported the AI generated a report in eight seconds that he described as accurate and better written than he could have produced manually.

Reported impact

Legal scholars and police watchdogs have raised concerns about relying on AI to generate documents that play a foundational role in investigative and legal processes, and that can influence who gets prosecuted or imprisoned. Andrew Ferguson, an American University law professor, expressed concern that the ease of the technology might cause police officers to be less careful with their writing.

Index analysis

Generative AI systems are prone to hallucinationโ€”fabricated facts or incorrect information in synthetic text. Axon stated it has reduced the GPT-4 'creativity dial' in Draft One to limit potential for hallucination, but questions remain about whether such controls adequately address the risks of using AI in reports that carry legal consequences.

Governance controls

Axon CEO Rick Smith acknowledged that district attorneys want to ensure officers, not solely AI chatbots, are responsible for authoring reports because officers may need to testify in court about what they witnessed.

Classification

Category
Reliability and hallucination
Industry
Government, security and defence
Country
United States
Harm status
Potential harm
Intent
Unintentional
Timing
Unknown
Harm types
Public interest
Harmed
General public
Failure modes
Hallucination
Consequences
None recorded
AI systems
Draft One

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

The Independent
Police officers are starting to use AI chatbots to write crime reports. Will they hold up in court?
26 August 2024
futurism.com
US police departments testing AI-generated incident reports
philasun.com
US police departments testing AI-generated incident reports