TL;DR: The UK government announced £4 million to build an AI-powered "interactive crime map" to predict criminal activity before it happens. Home Secretary Shabana Mahmood explicitly compared the vision to Jeremy Bentham's Panopticon, a prison where inmates are watched constantly. Prototypes are expected by April 2026, with full deployment by 2030. Critics call it a "frightening expansion" of surveillance that undermines the presumption of innocence. Meanwhile, a police chief just resigned after his force used Microsoft's AI chatbot to justify a policy based on an event that never happened.
The Home Secretary Said The Quiet Part Out Loud
Usually, government officials try to avoid dystopian framing when selling surveillance programs. Not Shabana Mahmood.
The UK Home Secretary explicitly invoked Jeremy Bentham's Panopticon, an 18th-century prison design where guards could observe all inmates without them knowing whether they were being watched. The psychological effect of potential constant surveillance was the point.
"The eyes of the state can be on you at all times," Mahmood said, advocating for AI-enhanced surveillance across England and Wales.
Philip K. Dick wrote about this in 1956. Spielberg made it into a movie in 2002. Now it's UK government policy.
What They're Building
The initial £4 million investment is part of the larger £500 million R&D Missions Accelerator Programme. Here's what's planned.
- Interactive crime map: Real-time tracking and prediction of theft, anti-social behavior, knife crime, and violent crime across England and Wales
- Data fusion: Combines information from police, local councils, and social services, including criminal records, previous incident locations, and "behavioral patterns of known offenders"
- Video search AI: Tool that analyzes CCTV, mobile phone footage, and doorbell cameras 60% faster, already tested by two unnamed police forces
- Risk identification: Officials specifically mentioned identifying "the 1,000 most dangerous men who pose the highest risk to women and girls"
Prototypes by April 2026. Full operational system by 2030.
College of Policing CEO Sir Andy Marsh laid out the strategy: "Test them with rigorous evaluation, and then spread them like wildfire through policing." Police chiefs are currently evaluating around 100 AI projects out of 1,400 "innovative practices."
The Problems They're Ignoring
Conservative MP David Davis cut through the PR: "If an AI system deems you to be at risk of committing a crime, how do you go about proving the AI is wrong?"
This is the fundamental problem with predictive policing. You're not responding to crimes. You're preemptively targeting people based on algorithmic guesses about future behavior.
Davis also pointed out that predictive algorithms create a "postcode lottery of justice": if you live in an area with historically high crime rates, the AI will flag you more often. Not because you've done anything, but because of where you live and who your neighbors are.
Meanwhile, 94% of crimes in England and Wales go unsolved. The police aren't solving existing crimes. They want AI to predict new ones.
Civil Liberties Groups Sound the Alarm
Big Brother Watch called the plan "a frightening expansion" of surveillance powers, warning that:
- It undermines the presumption of innocence
- It expands state monitoring into new areas
- It embeds historical policing bias into algorithmic decisions
That last point is critical. The AI learns from historical policing data. If police have historically over-policed certain communities, the AI will identify those communities as higher risk. Bias in, bias amplified out.
Baroness Shami Chakrabarti, founder of Liberty, described facial recognition and predictive technologies as "incredibly intrusive" and warned deployment moves the UK closer to "a total surveillance society." She stressed the urgent need for legal frameworks before deployment, frameworks that don't exist.
The AI Already Failed Spectacularly
Here's the timing: The UK government announced this predictive policing expansion the same month that Craig Guildford, chief constable of West Midlands Police, resigned over an AI failure.
The force banned Israeli fans from a football match between Maccabi Tel Aviv and West Ham. Their justification? Intelligence about violent incidents at previous matches.
The problem: The "intelligence" came from Microsoft's Copilot AI chatbot, which hallucinated a violent match that never happened. The police initially claimed they used Google searches. Then they admitted it was an AI chatbot making things up.
This is the foundation they want to build predictive policing on.
The US Tried This. It Didn't Work.
Predictive policing isn't new. US cities have experimented with it for years. The results aren't encouraging.
PredPol (now Geolitica) was one of the most widely deployed systems. Studies found it didn't reduce crime rates better than existing methods. It did increase police presence in low-income and minority neighborhoods. Several departments abandoned it after community backlash.
The RAND Corporation found predictive policing tools showed "limited effectiveness" and raised significant civil liberties concerns. The Brennan Center documented how these systems consistently flagged the same neighborhoods regardless of actual crime trends.
The UK is repeating mistakes the US already made.
What You Can Do
Contact Your MP
The Technology Secretary and Home Office are pushing this without meaningful parliamentary debate. Your MP can raise questions. WriteToThem.com makes it easy.
Support Civil Liberties Groups
Big Brother Watch, Liberty, and the Open Rights Group are actively fighting these expansions. They need funding and public support.
Follow the Implementation
April 2026 prototypes mean this is happening now. Watch for pilot programs in your area. Document police AI use. Request information under FOI about what your local force is testing.