Scotland Yard announced this week that live facial recognition cameras will cover London's West End by December — Soho, the theatres, the retail strips. Britain is not sleepwalking into a surveillance state. It is sprinting.
The scale of what is actually being built right now
I think the critics are right, and I think the government knows it. Home Secretary Shabana Mahmood called this the "most significant modernisation of policing in nearly 200 years." That framing is doing a lot of work. What she is actually describing is a national biometric dragnet, dressed up in the language of reform.
The numbers are not abstract. The Home Office has pledged 40 new live facial recognition vans to roll out across town centres in England and Wales. On top of that, £115 million over three years goes into a new National Centre for AI in Policing, branded Police.AI. Police chiefs are already evaluating around 100 AI projects. This is not a pilot. This is infrastructure.
And yet the legal framework to govern any of it does not exist yet. No laws in the UK currently mention facial recognition, and the technology has never been formally debated by MPs. A public consultation ran for ten weeks and closed in February 2026. Deployment continues regardless.
Palantir is already inside the machine
The facial recognition story is alarming on its own. But it sits inside something larger and far less discussed: the quiet embedding of Palantir across every layer of British public life. The US firm now holds at least £670 million in UK government contracts spanning defence, health, policing, and critical infrastructure.
Two senior systems engineers at the Ministry of Defence broke cover to warn that Palantir poses "a national security threat to the UK." Their core argument is not about data ownership. It is about the mosaic effect: individually harmless data points, combined, reveal state secrets. One source described how three pieces of unclassified logistics data could expose the location of a nuclear submarine on a specific date.
“Palantir does not need to own the data. They can extract, transform and exploit the metadata to build their own rich picture.”
— Anonymous MoD systems engineer, via The Nerve
The government's defence is contractual guarantees. Defence minister Luke Pollard assured the Commons that MoD data "will remain under the ownership of the MoD." This is unserious. Ownership of raw data is irrelevant when a private company controls the analytical engine that turns that data into intelligence.
The strongest counterargument, and why it falls short
I remember watching the footage from the Southport riots in August 2024 and thinking: yes, we need tools that work fast. Retrospective facial recognition identified 127 suspects from that disorder. A six-month trial in Croydon produced more than 170 arrests. The Met says Live Facial Recognition has contributed to over 2,000 arrests since 2024, including suspected rapists and domestic abusers.
That is the good edge, and it is real. Policing minister Sarah Jones has compared facial recognition's impact to the DNA breakthrough of the 1980s — and that comparison is not entirely wrong. Speed matters when a violent offender is at large.
But here is where the argument collapses. Essex Police paused its own live facial recognition deployments after identifying accuracy and bias risks. The Home Office itself acknowledged historic bias in the algorithm used for retrospective searches on the Police National Database. The Baroness Casey Review found institutional racism inside the Metropolitan Police — the largest user of this technology in the country. Handing a biased institution a biased tool and calling it reform is not a solution.
Your passport photo is already in a police lineup
The scope of the database question is the part everyone skips. Freedom of Information requests revealed that the government covertly allowed police to search 150 million UK passport and immigration photos for facial recognition matches. You did not consent to that when you applied for a passport. Campaign groups are now preparing a legal challenge.
Meanwhile, Merseyside Police officers now carry handheld facial recognition apps. West Midlands Police is running real-time facial recognition from body-worn cameras. At least one force has discussed using drone-mounted cameras to identify people at protests. Would you trust a system this expansive, governed by rules not yet written, operated by forces with documented discrimination problems?
The NHS angle makes this worse. Palantir's Federated Data Platform now runs across NHS England. Palantir UK CEO Louis Mosley has said that if a Reform government comes to power, the company will follow directives to use NHS data to target individuals based on immigration status. That is not a hypothetical. That is a stated position from the company currently holding your health records.
The Commons Science, Innovation and Technology Committee put it plainly: reliance on a small number of US-based providers represents a "clear vulnerability." The Swiss army rejected Palantir over fears that sensitive data could be accessed by US intelligence services. Britain, by contrast, gave Palantir a £1.5 billion investment deal and a tour of its defence ministry.
This is cowardly governance dressed as boldness. The EU AI Act restricts live facial recognition for law enforcement. Some US cities have banned it outright. Britain is standardising, expanding, and connecting these systems into a national network before the legal framework exists to constrain them. The infrastructure will outlast any government that builds it.