The UK’s Independent Office for Police Conduct has launched an independent investigation into a senior Derbyshire Constabulary detective accused of using Microsoft Copilot to make decisions in criminal cases. According to the BBC, the officer allegedly used the AI assistant to formulate policy decisions, assess and manage risk, and justify decision-making — and entered sensitive operational and personal case data into the tool. Derbyshire had authorised Copilot for officers in May 2025 strictly for drafting reports and summarising information, with an explicit ban on inputting sensitive data. The officer has been served a gross-misconduct notice, though the IOPC stresses no finding has been made.
The case reached the watchdog after Derbyshire Constabulary made a voluntary conduct referral on June 23, following an internal review of the officer’s AI use. The IOPC assessed the referral and decided the matter required an independent investigation rather than one handled by the force itself — a signal of how seriously it views the allegations. The force itself had authorised Copilot, which puts the institution in the awkward position of policing the boundaries of a tool it introduced. Emily Barry, the IOPC’s director of engagement, called them “serious allegations” that deserve thorough and independent investigation.
Investigators will examine the detective’s Copilot use between May 2025 and June 2026 to determine whether professional standards were breached. The allegations, as set out by the IOPC, are that the officer used AI to create policy decisions, assess and manage risk, and justify decision-making — and that sensitive operational and personal case data was inputted into the tool, in direct breach of the force’s conditions.
The Rules Copilot Was Supposed To Follow
Derbyshire Constabulary authorised Microsoft Copilot for officers in May 2025 as a productivity aid: drafting reports, summarising information, the routine paperwork of policing. The permission came with a hard condition — no operational, sensitive or personal data was to be entered into the AI. The detective is accused of breaching both the spirit and the letter of that guidance: using the assistant not merely to handle case material, but to make and justify the decisions themselves. The case now tests what happens when that line is crossed, and whether internal guidance is enough to govern general-purpose AI tools inside institutions that handle the public’s most sensitive information.
Why It Matters
This is one of the first formal misconduct investigations anywhere centred on an officer’s use of a mainstream AI assistant — and it lands just as governments wrestle with where AI belongs in public service. Sweden has just blocked ChatGPT and Claude from government computers; the US has purged DeepSeek from civilian agencies. The Derbyshire case shows the other side of the same problem: even where AI is officially permitted, the boundary between acceptable assistance and unacceptable delegation is thin, and crossing it can now end a career.
The case also raises an uncomfortable question for Microsoft and its enterprise customers: when a general-purpose assistant is embedded in high-stakes workflows, whose responsibility is it when users push it beyond its brief? Feeding live case data into a cloud AI service echoes the Samsung engineers who pasted confidential code into ChatGPT in 2023 — except here the data belongs to criminal investigations, and the stakes include people’s liberty.
Background
Police forces have been early and sometimes careless adopters of AI paperwork tools. US departments have used Axon’s Draft One to generate report drafts from body-camera audio, and earlier this year an AI-generated report in Utah claimed an officer had “turned into a frog” after picking up audio from a Disney film playing in the background. The IOPC itself warned earlier this month that officers risk “accidental” AI misconduct — a warning the Derbyshire investigation has now made concrete.
Source: TechSpot (reporting the IOPC/BBC)
0 Comments