AI surveillance trial costs £320k, catches zero crooks

Our Take

Credit to Pivot to AI for doing the unglamorous work of actually checking whether a hyped surveillance system did what it promised. Spoiler: it didn't. A £320,000 AI policing trial in the UK ran, scanned, flagged, and ultimately caught nobody. That's not a rounding error — that's the entire business case for the technology collapsing under its own numbers.

This matters here because the same pitch is being made to city councils across the US right now: give us the budget, trust the vendor's marketing deck, and public safety will follow. Flock Safety and its ALPR network have expanded into thousands of communities on almost identical promises, with almost as little independent auditing of hit rates, false positives, or actual case outcomes. When the data does surface, it tends to look a lot like this UK trial — lots of plates scanned, lots of driver movement logged, very little evidence it's solving or preventing the crimes it was sold to stop.

The lesson isn't that AI surveillance never works, it's that nobody should take a vendor's word for it, and nobody should write the check without demanding the audit first. If there's a camera network in your town, check our map to see what's already watching your block, and if you want to push back on a deployment or budget vote, our take-action page has the tools to start asking your city the same hard questions this video asked the UK police.