A criminal investigation in England has put a sharp, real-world edge on a question the AI world usually debates in the abstract: what happens when machine-generated content reaches the justice system?

What the force has confirmed

Derbyshire Constabulary has confirmed it has launched a criminal investigation into an allegation of perverting the course of justice, after an officer allegedly used AI systems to create 'evidential material' in a number of cases. The officer has been removed from frontline duties pending the outcome.

The force says the investigation is at an early stage, that no arrests have been made, and that it is working closely with the Crown Prosecution Service (CPS) in relation to any cases that may have been affected. That last point matters: it means real prosecutions could now be reviewed for whether the material in them can be relied on.

The important caveats

This is an allegation, not a finding. Nothing has been proven, no one has been charged, and the officer has not been named. We are reporting that an investigation exists and what the force has said about it — not a verdict. Perverting the course of justice is a serious offence precisely because the integrity of evidence is the foundation the whole system stands on, which is also why an allegation like this is handled carefully and slowly.

Why it matters beyond one case

The detail that makes this more than a local crime story is the phrase 'evidential material.' Generative AI can now produce convincing text, images, audio and video on demand. The same capabilities that power a chatbot or an image generator can, in the wrong hands, manufacture things that look like documents, statements or records. The justice system was not built with that in mind — it largely assumes that a document is what it claims to be.

It also lands at an awkward moment. UK police forces are actively piloting AI tools to help write reports and witness statements — there are live evaluations of exactly this, run with universities, aimed at saving officer time. Those pilots are legitimate and tightly framed. But a case like this underlines why the guardrails around them — what AI may touch, what it may not, and how anything it produces is logged and disclosed — have to be unambiguous. The line between using AI to draft a report faster and using AI to create something that is then presented as evidence is the entire ballgame, and it needs to be visible from the outside, not just trusted.

For now, the facts are limited and the process is just beginning. We will update this as Derbyshire Constabulary and the CPS say more. But the marker it puts down is already clear: as generated content gets cheaper and better, every institution that runs on the authenticity of a document — courts first among them — needs a way to prove what is real.