UK police watchdog warns facial-recognition checks fall short, risking wrongful arrests

Innocent people, especially those from ethnic minorities, risk being wrongly arrested when police act on facial-recognition matches without enough checks. A new UK police watchdog report found that some investigators mistook search results for evidence, while biased algorithms and unsuitable photographs added to the danger of identifying the wrong person.

His Majesty’s Inspectorate of Constabulary published the report on Wednesday, according to the Guardian. It examined retrospective facial recognition: searches using existing photographs, rather than live systems scanning people in real time. Its findings concern both the reliability of potential matches and what officers do after receiving them.

A possible match is not evidence

The report found that some investigators did not understand that retrospective facial-recognition results were intelligence requiring further inquiries, rather than evidence ready to support enforcement action. That distinction puts a clear responsibility on investigators: a suggested identity needs corroboration before officers act on it.

Inspectors found that some officers had not made enough corroborative inquiries before acting on potential matches. In a small number of cases, excessive reliance on the results led to inappropriate investigative action. The report warned of incorrect identification or arrest; those findings should not be read as a count of confirmed wrongful arrests.

Pete Fussey, a University of Southampton professor and former police adviser on facial recognition, said the examples showed investigators accepting computer suggestions without further checks. He described meaningful human scrutiny of AI systems as a legal requirement—not simply an optional precaution when an officer doubts a result.

Some investigators didn’t know that RFR results were intelligence rather than evidence and required further enquiries before officers took enforcement action.

His Majesty’s Inspectorate of Constabulary report, quoted by the Guardian

Bias and poor images weaken the search

The watchdog also found bias in the algorithm, which was more likely to wrongly flag people from ethnic minorities as suspects. That makes the failure to check results particularly consequential: the risk of a mistaken suggestion was not evenly distributed across the people who could become subjects of an investigation.

Auditors identified a separate problem with the photographs submitted for searches. Users uploaded blurred pictures and images containing multiple faces, among other unsuitable material. These inputs reduced the system’s ability to return reliable matches, adding an image-quality problem to the report’s concerns about algorithmic bias and investigators’ decisions.

Retained photographs and an unfinished legal framework

Inspectors found that custody photographs that should have been deleted were still being kept. The report identified that retention as another risk to public confidence, separate from whether a particular search produced an accurate match. Its concerns therefore extend to the handling of police-held images, not just search performance.

The report said legislation had not kept pace with facial-recognition technology and there was no comprehensive statutory framework governing its use. That gap made it harder for police forces to understand both the lawful basis for retrospective searches and the safeguards needed to protect the public.

Inspector Kathryn Stone called for a clearer legal framework, stronger national standards and better governance and oversight. She said forces must demonstrate that safeguards exist, bias and data-retention risks are managed, and decisions remain subject to human judgment. Those protections, she said, are essential to maintaining public confidence.

Police leaders defended the technology’s investigative role. Ben Russell, the National Police Chiefs’ Council lead on facial recognition, said policing remained committed to its ethical, proportionate and responsible use. He welcomed the watchdog’s recognition of its possibilities for investigating crime, bringing offenders to justice and keeping communities safe.