Police in England and Wales Ordered to Halt AI Use in Court Statements
Police forces in England and Wales have been directed to cease the use of artificial intelligence in the preparation of court statements. This intervention follows discoveries that commercial AI tools were being deployed without proper assessment and that reliance on human review was insufficient to prevent errors in legal documentation.
Failure of AI Governance and Oversight
Commercial AI tools were deployed within police forces before undergoing rigorous assessment. The current governance model relied heavily on a "check everything it produces" policy, which critics argue is an ineffective safeguard against the inherent risks of generative AI.
One observer noted that the reliance on human review is not a viable solution, stating:
"All forces will have a policy that says, ‘Check everything that it produces’... obviously isn't an actual solution here, of course people will get complacent and throw stuff over the wall whenever they can."
Risks to Legal Integrity and Evidence
The use of AI in court statements introduces significant risks regarding the accuracy and authenticity of legal testimony. Because the specific wording used to describe an event is critical to legal proceedings, the substitution of an officer's original phrasing with AI-generated text may compromise the integrity of the evidence.
Concerns have been raised regarding the distinction between boilerplate text and subjective information. While AI might handle routine templates, subjective descriptions of events must originate from the officer's own observations to remain probative in court. The risk of "hallucinations"—where AI generates false but plausible-sounding information—is particularly dangerous in a judicial context where testimony is used to determine guilt or innocence.
Proposed Alternatives to AI-Generated Text
Rather than using AI to draft text, some suggest moving toward more probative forms of initial reporting to save labor while increasing reliability. A proposed alternative is the use of immediate video narration of facts and decisions following an incident.
This approach offers several advantages over AI-generated text:
- Tamper Evidence: Video is considered more resistant to AI manipulation than plain text.
- Freshness: Oral reports captured immediately after an event preserve details that may be lost in delayed written reports.
- Context: Video captures body language and tone, providing more information than a transcribed or AI-summarized report.
Broader Implications of AI Productivity Pressures
The push to integrate AI into police work is seen by some as part of a larger trend of government spending aimed at closing "productivity gaps" through expensive commercial AI contracts. There is skepticism regarding whether these tools actually fill the productivity void or simply introduce new risks into critical public infrastructure.