Should immigration authorities be allowed to scan people's faces in public? I've been reading about the UK's plans to expand live facial recognition for immigration enforcement. The Home Office has already been testing t…
두번클릭@twoclicks_later
It is easy to feel reassured when facial recognition comes with the explanation that "a human performs the final check." However, it would be better to start by distinguishing what exactly the person is checking. Verifying that the faces match is a different task from determining whether that person should be sought at that moment.
The 2020 Bridges appellate court ruling in the UK helps in considering this distinction. At the time, the court found issue with the South Wales Police's facial recognition operations because they had too much discretion regarding who was placed on the watchlist and where cameras were placed. It also noted that the mere mechanism of a human checking after an alert was insufficient to fulfill the duty to investigate potential gender and racial biases. This should be distinguished from a conclusion that actual bias was proven. Furthermore, this ruling cannot be applied to judge the overall legality of immigration enforcement in 2026.
Therefore, I suggest explaining the review process in three stages. Before deployment, the location, time, purpose, and conditions for inclusion on the watchlist are reviewed. When an alert occurs, the facial match and the validity of the underlying records are verified separately. Finally, procedures are set for who stops the process and re-evaluates when an individual raises an error regarding the records. This is not a description of how the UK currently operates, but a set of procedures I believe should be required when discussing an expansion.
For example, one could assume a case where a face matches accurately, but an outdated or corrected record remains on the list. Simply comparing the photo one more time would not verify the validity of the record itself. I believe that if the scope is expanded to include immigration enforcement, the nature of the violation and the reasons why other means are difficult to use must also be explained separately.
There is a counterargument that increasing the number of verification steps could slow down responses in urgent situations. Human facial verification remains necessary. My suggestion is to review the location and target criteria before deployment, and to predetermine the records to be checked and who is responsible for stopping and re-examining the process on the ground. It has not been proven that this alone reduces delays or errors.
I believe what should be made public is not the list of people's names or faces, but the criteria and scope of responsibility at each stage. No matter how carefully on-site verification is performed, it does not automatically justify the purpose of the deployment or the conditions for watchlist inclusion. Only by distinguishing what reviews have been completed and what remains can we evaluate the claim that "a human performs the check" as an actual safety measure.
Reference: Bridges ruling paragraphs 120-132, 181-191