China Facial Recognition Rules: UK Scanned 4M Faces First

Picture yourself walking through a shopping street on a Saturday afternoon. In under five hours, a single police camera in central London scanned fifty thousand faces. Yours could have been one of them, and you'd never have known.

Here's why this touches everyone, not just people

Here's why this touches everyone, not just people who follow policing. If you've ever walked past a parked police van in a busy town center, your face may already have been measured, compared, and checked against a list. In 2025, U.K. police scanned around four million faces this way. Now the government wants to grow its fleet of mobile scanning vans, from about ten to fifty. And the plan is to spread them from a handful of forces to all forty-three across England and Wales. So the real question isn't whether this technology works. It's whether the rules have caught up with how fast it's rolling out.

Let's start with the vans themselves. Devon and Cornwall Police recently ran their first mobile deployment in the seaside town of Newquay. Officers say if your face isn't on their watchlist, the system deletes your image within seconds. That sounds reassuring, and it's meant to. But Newquay isn't a one-off. It's the template. Police forces across the country are copying that exact playbook.

Now, the accuracy question, because this is where things get uncomfortable. Police point to testing from the U.K.'s National Physical Laboratory. That testing shows these systems are correct about ninety-nine percent of the time. Sounds airtight. But that number describes a very different job. It's a lab comparing a still photo against a database, quietly, under controlled conditions. That's not what happens on a crowded street.

Here's the number that actually matters

Here's the number that actually matters. According to the government's own figures, Metropolitan Police cameras produced over two thousand alerts across a run of deployments. Of those, more than two hundred were false, the system flagged the wrong person. That's nearly one in every ten alerts, pointing at someone who did nothing wrong. Think about that on a busy pavement. It could be you being stopped because a camera made a bad guess.

And the errors aren't random. Police themselves acknowledge these systems can carry gender and ethnic bias. Essex Police only paused their operations after those concerns surfaced. For anyone worried a machine might misjudge them because of how they look, that worry has evidence behind it.

There's one more piece. All of this expansion happened before any dedicated law existed to govern it. Only in 2026 did the government announce a proper legal framework, years after the vans were already on the streets. Deploy first. Write the rules later.