UK immigration enforcement publishes updated facial recognition policy, with Manchester deployments listed
Version 2.4 keeps children off watchlists and bars action on an alert alone. Manchester city center deployments are listed for November 25 and 26.
By AI Desk ·
Britain's Home Office has published the latest version of the rules its Immigration Enforcement arm follows when it scans faces in public places, and the unit has listed its next deployments, in Manchester city center on November 25 and 26, 2026. The policy, published on November 18, is version 2.4: its history table starts with a first version in October 2025 and ends with a change to the oversight wording on November 3, 2026. The Manchester dates were added to the Home Office's live facial recognition page on November 19 and were still listed when we checked at 13:05 UTC on November 22.
Immigration Enforcement has used the technology before, at Holyhead Port. The Home Office's published results cover sessions there in November 2025 and February 2026, and its page records a trial at the port in the week beginning February 23, 2026.
Who can be on a watchlist
Watchlists are limited to people the unit lawfully seeks for a set purpose who cannot reasonably be found by less intrusive means, and who fit set categories. For criminal enforcement, those are people under a deportation order, wanted on a UK court warrant for an immigration-related offense or sought on a charge over one; for general enforcement, people liable to detention, actively sought for enforcement or removal, or in breach of their conditions of stay or removal directions.
Nobody under 18 may be included, and nobody may be added on the basis of intelligence interest without objective evidence, association, speculative risk or general monitoring alone.
Safeguards during and after a deployment
Faces are compared only with the authorized watchlist; trained staff review any alert and officers decide whether to act. An alert alone never leads to enforcement action, people are not required to cooperate with having their image captured, and declining to engage is not in itself a reason to act. Images that trigger no alert are deleted automatically and permanently, and data linked to a false alert within 24 hours.
Each deployment needs written authority from an officer of at least Grade 7. Higher-risk deployments, which include those at high-footfall locations, outside ports or other controlled environments, or in places with heightened community sensitivity, also need recorded oversight at deputy director level. The Home Office page does not say how it classes the Manchester deployments; on our reading of those definitions, a city-center deployment away from a port falls into the higher-risk group. Where it would not undermine an operation, the unit is to give public notice on Home Office websites at least five days ahead.
What the Holyhead records show
| Period | Entries | Watchlist size | Faces seen | Alerts | Incorrect alerts | Arrests or disposals |
|---|---|---|---|---|---|---|
| November 10-12, 2025 | 1 | 1,942 | 2,038 | 0 | 0 | 0 |
| November 27-29, 2025 | 7 | 4,867 to 4,868 | 5,474 | 2 | 0 | 1 |
| February 24-26, 2026 | 10 | 6,535 | 2,752 | 0 | 0 | 2 |
| Total | 18 | - | 10,264 | 2 | 0 | 3 |
Every entry used a threshold of 0.64, the similarity score at which the system raises an alert. The policy says a count of faces scanned, which we take to match the table's faces seen, records each time the software identified a face and checked it against the watchlist, not the number of people. By our reading, two February entries each record a positive intervention and an arrest or disposal without an alert; the table does not explain them.
On accuracy, the policy relies on a National Physical Laboratory study of police live facial recognition at five deployments in London and Cardiff in July and August 2022. As the policy reports it, the study found no statistically significant gender or race differences in correct matches at a threshold of 0.60, though correct matches rose with age; false matches were rare at 0.60, 0.62 and 0.64, with no significant imbalance between groups, and there were none at 0.64 or above. Immigration Enforcement says it will keep using 0.64.
The legal basis
The document says it creates no new powers: facial recognition helps identify people for powers Immigration Enforcement already has. Data processed for detention and removal work falls under the UK GDPR and Part 2 of the Data Protection Act 2018, and data for criminal work under Part 3. Each deployment must record the safeguards that keep any interference with privacy rights under Article 8 of the European Convention on Human Rights lawful, necessary and proportionate.
Objections, as the government summarizes them
This report does not quote critics of the policy directly. The objections here come from the Home Office's November 4 summary of responses to a separate consultation on facial recognition in law enforcement, covered in our November 5 report. The summary is the government's account of respondents' views, not their own words, and does not mention Immigration Enforcement. According to it, some feared facial recognition could grow into routine mass surveillance without binding limits, others said innocent people should not routinely be put on watchlists, and some wanted live deployments kept temporary and tightly defined. Respondents included members of the public, civil liberties groups, police bodies and regulators.
Sources
- Immigration Enforcement: Live Facial Recognition Policy Document Home Office and Immigration Enforcement (GOV.UK) gov.uk
- Live facial recognition in Immigration Enforcement Home Office and Immigration Enforcement (GOV.UK) gov.uk
- Live facial recognition deployment results (accessible) Home Office (GOV.UK) gov.uk
- New legal framework for law enforcement use of biometrics, facial recognition and similar technologies: government response to the consultation (accessible) Home Office (GOV.UK) gov.uk
Checked against the 4 sources above on November 22, 2026. Spotted an error? Tell us.