Police watchdog warns facial recognition errors risk wrongful arrests

Inspectors at His Majesty's Inspectorate of Constabulary and Fire & Rescue Services published the findings on Wednesday, covering only retrospective facial recognition — the after-the-fact searches used to hunt suspects, not the live real-time scanning that draws most protests. The problem they found is basic: some investigators did not realise RFR results are intelligence, not evidence, and that more enquiries are required before officers act. That gap, the report says, means people could be wrongly identified or arrested, and public trust in the tech could take the hit.

The algorithm's bias problem hasn't gone away. The report found it is still more likely to wrongly flag people from ethnic minorities as suspects, and auditors spotted users uploading junk probe images — blurred shots, pictures with several faces in them — which drag down the system's reliability. Inspectors also found custody images that should have been deleted were kept, and said legislation has not kept pace with the technology, leaving no comprehensive statutory framework for how forces may lawfully use it.

His Majesty's Inspector of Constabulary Kathryn Stone said public backing depends on showing the tools are used properly. Forces, she said, must prove safeguards are in place, that bias and data retention risks are managed, and that a human still makes the call — adding that a clearer statutory framework, stronger national standards and better oversight are essential to keeping confidence.

Jasleen Chaggar of Big Brother Watch called the findings alarming, saying officers treating a probability score as bona fide evidence sharply raises the odds of innocent people getting a knock on the door over crimes in towns they have never visited. She also flagged that police are not even logging their use of retrospective facial recognition, meaning people can become suspects without knowing why, and warned against running a biased algorithm against millions of unlawfully held custody images.

Ben Russell, the National Police Chiefs' Council's lead on facial recognition, said policing stays committed to ethical, proportionate and responsible use of the technology to investigate crime and bring offenders to justice, and welcomed the inspectorate recognising its possibilities. Prof Pete Fussey of the University of Southampton, a former police adviser on the technology, pointed out that meaningful human scrutiny of AI systems is a legal requirement — yet the report shows investigators taking the computer's suggestions at face value.