Skip to content
Submit incident
AII-1002-03347
Well documentedThree or more independent publishers report the core facts.

Axon Draft One AI police report system lacks transparency mechanisms

Axon Enterprise's Draft One system automatically generates police reports from body-camera audio using AI, but reportedly lacks transparency mechanisms that would allow auditing or identification of AI-generated content in reports. An Electronic Frontier Foundation investigation found the system was deliberately designed to prevent such audits, raising accountability concerns in the criminal justice system.

10 July 2025 Β· Axon Enterprise Β· Governance, accountability and labour
10 July 2025
AII-1002-033477 sources ↓
Quick facts
Category
Governance, accountability and labour
Country
United States
Source
Ars Technica
Date
10 July 2025
Drafted by
Raidu Incident Detection

What happened

Axon Enterprise developed Draft One, a generative AI product that automatically writes police reports based on audio from officers' body-worn cameras using a ChatGPT variant.

An Electronic Frontier Foundation investigation found that the system is deliberately designed to prevent auditing. It is often impossible to tell which parts of a police report were generated by AI and which parts were written by an officer, and the product itself does not allow agencies to isolate reports containing AI-generated narratives.

Reported impact

The lack of transparency undermines accountability in the criminal justice system and prevents meaningful audit of Draft One usage by police chiefs, researchers, or the public.

In jurisdictions like California, police narrative reports are already exempt from public disclosure, while the technical design of Draft One compounds these barriers by providing no way to identify which reports contain AI-generated content.

Index analysis

The system uses bracketed placeholders that officers are encouraged to add to or delete from, which can obscure the AI's role in generating report text.

Axon's user manuals do not provide clear documentation of logs or settings that would enable external parties to determine how and when the AI was used in individual reports.

Governance controls

Police agencies can voluntarily institute measures to make AI-generated reports searchable by keyword, but no mandatory disclosure mechanism exists.

The system settings for Draft One contain multiple levers that police administrators can adjust to change how and when officers use the software, but no public oversight of these settings has been established.

Classification

Category
Governance, accountability and labour
Industry
Government, security and defence
Country
United States
Harm status
Potential harm
Intent
Intentional
Timing
Pre-deployment
Harm types
Public interest
Harmed
General public
Consequences
None recorded
AI systems
Axon Draft One

Sources and evidence

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

Ars Technica
Cops' favorite AI tool automatically deletes evidence of when AI was used
10 July 2025
StateScoop
AI tool that writes police reports needs better oversight, transparency, report says | StateScoop
10 July 2025
Electronic Frontier Foundation
Axon's Draft One is Designed to Defy Transparency
10 July 2025
The National Desk
Time saved or justice threatened? Police departments are using AI to write reports
11 July 2025
Democratic Underground
A new AI tool to help police write reports deliberately hides who - officer or chatbot - wrote what in the report
11 July 2025
eff.org
Axon Draft One AI police report system lacks transparency mechanisms
eff.org
Axon Draft One AI police report system lacks transparency mechanisms