On 21 April 2026, a Divisional Court of the High Court — Lord Justice Holgate and Mrs Justice Farbey — dismissed the judicial review brought by anti-knife campaigner Shaun Thompson and Big Brother Watch director Silkie Carlo against the Metropolitan Police's live facial recognition (LFR) programme. The court held the Met's 2024 LFR policy compatible with Articles 8, 10, 11 and 14 of the European Convention on Human Rights, finding it contained "adequate and lawful constraints on where LFR may be deployed" and gave the public "an adequate indication of the circumstances in which LFR will be used" (R (Thompson and Carlo) v Commissioner of Police of the Metropolis [2026] EWHC 915 (Admin)).
Thompson's complaint was not abstract. In February 2024, near London Bridge, the system matched his face to his brother's image on a watchlist. He was stopped, held for roughly half an hour, repeatedly asked for fingerprints and threatened with arrest before officers accepted he was not the man they wanted. The Met later paid him a settlement on his damages claim. He and Carlo, who called the ruling "disappointing" and the "fight far from over," intend to appeal.
The strongest case for the Met
Begin where the police are on solid ground. The operational record the Met put before the court is genuinely impressive, and it would be dishonest to wave it away. According to the Home Office's December 2025 factsheet, LFR deployments in London from January 2024 to September 2025 produced over 1,300 arrests for serious offences, plus more than 100 arrests tied to monitoring registered sex offenders. The accuracy figures are far from the dystopian caricature: an 89% chance of correctly flagging someone actually on the watchlist, and a worst-case false-match rate of roughly 1 in 6,000 against a 10,000-image list — with the real-world false-alert rate better still. Figures cited around the trial put scans at around 3 million faces in a year against only 12 recorded false alerts. The Met's current policy is also materially tighter than the regime that was struck down in 2020. That is the case for the technology, and it is a serious one.
What the court actually decided
The judgment is narrower than the headlines suggest, and the narrowness matters. The claimants framed their challenge tightly, asking whether the policy was "in accordance with the law" — accessible, foreseeable and non-arbitrary. The court answered that question, and only that question. It found the policy's "clear, interlocking and cumulative constraints" on use cases, watchlists, locations and senior authorisation meant deployment decisions were governed by rules rather than officer "whim."
What the court did not decide is just as important. As legal analyst William Page noted for Inforrm, the bench did not rule on whether any given deployment was necessary or proportionate in practice — only whether the policy was foreseeable in principle. Evidence that identical watchlists were reused across different locations "fell outside the scope" of the judgment. The prospect of permanent, fixed LFR cameras was set aside as "speculation." And on discrimination under Article 14, the court left the door open: a properly evidenced bias claim might succeed, but this one was not made out. This is a ruling about a document, not a blank cheque for a surveillance architecture.
Bridges said this needed a law. It still doesn't have one.
The deeper problem is what the case reveals about Britain's approach to a powerful technology: there is still no bespoke statute governing police facial recognition. Deployment rests on general police powers, data-protection law, and forces' self-written policies — tested, if at all, by occasional judicial review.
That is precisely the gap the Court of Appeal flagged six years ago. In R (Bridges) v Chief Constable of South Wales Police [2020] EWCA Civ 1058, the court found that force's facial-recognition trials unlawful, citing "fundamental deficiencies" in the legal framework that left too much discretion to individual officers over where and how to deploy, alongside failures under the Data Protection Act 2018 and the public sector equality duty. The clear implication was that Parliament needed to legislate. Six years on, it has not. The Home Office only closed a consultation on reforming biometrics law in February 2026.
Scaling before the rules are settled
Meanwhile, the technology is being scaled aggressively. In January 2026, Home Secretary Shabana Mahmood announced the largest facial-recognition expansion in UK history: 40 new LFR vans, lifting the national fleet from 10 to 50, alongside £115 million over three years for a National Centre for AI in Policing. Thirteen forces had already used the technology by late 2025. The roadmap is national; the legal foundation remains a patchwork that a single appeal could unsettle.
The pro-innovation case is a statute, not a moratorium
None of this argues for banning LFR. Used well, it is a legitimate and effective tool, and the arrest figures are real. But effective tools deserve durable, legitimate foundations — and policy-by-litigation provides neither. A clear primary statute would set explicit thresholds for deployment, define and limit watchlist composition, mandate independent oversight and published accuracy and bias audits, and guarantee redress for people like Thompson. That is what "in accordance with the law" should mean in substance, not merely in the thin sense a sympathetic bench can certify today and a different one might question tomorrow.
Innovation thrives on legal certainty. A surveillance capability expanding to 50 vans nationwide on the strength of a self-authored policy is not certainty — it is borrowed time. Parliament was told to act in 2020. The Met's win in 2026 is the strongest argument yet that it finally should.