🔗 Share this article

Police forces across the UK successfully lobbied to use a facial recognition system known to be biased against females, youths, and individuals from ethnic minority groups, after complaining that a less biased version produced a reduced number of investigative leads.

The Technology in Practice

UK forces use the national police database to conduct retrospective facial recognition searches. This process involves comparing a reference photograph of a person of interest against a repository of more than 19 million mugshots to find potential matches.

Admitted Bias

The UK interior ministry admitted last week that the system was biased. This admission followed a study by the National Physical Laboratory (NPL) determined it misidentified people of Black and Asian heritage and females at much greater frequency than Caucasian males. The ministry said it “had acted on the findings”.

“It prompts the question of whether facial recognition only becomes useful if users tolerate biases in ethnicity and gender. Operational ease is a poor argument for disregarding fundamental rights.”

Long-Standing Problem

Official papers reveal that this discriminatory flaw has been known about for over twelve months. Furthermore, police forces lobbied to reverse an earlier ruling that was designed to mitigate the problem.

Police bosses were notified of the system's bias in September 2024. The government-ordered NPL review found the system was had a higher probability to produce false positives for photos of females, individuals of Black ethnicity, and those under 40 years old.

A Reversed Decision

In response, the National Police Chiefs’ Council (NPCC) ordered that the accuracy setting required for potential matches be raised to a point where the bias was greatly diminished.

However, this decision was reversed the following month following complaints from police that the modified technology was generating a lower number of “investigative leads”. NPCC documents show the higher threshold cut the proportion of searches resulting in possible identifications from 56% to a just 14%.

Severe Disparities

Although the authorities declined to specify what threshold is now in operation, the recent NPL study discovered the system could generate incorrect matches for women of Black heritage almost 100 times more frequently than for white women at certain settings.

The Home Office commented on these findings: “The testing identified that in a limited set of circumstances the software is more likely to incorrectly include some demographic groups in its search results.”

Operational Effectiveness vs. Bias

Outlining the impact of the brief increase to the system's accuracy setting, the police records note: “This adjustment greatly lessens the impact of discrimination across protected characteristics of ethnicity, age and gender but had a substantially detrimental effect on operational effectiveness”. The documents further note that police units argued that “a previously useful tool returned results of limited benefit”.

Wider Implementation Proposals

Meanwhile, the government has launched a two-and-a-half-month consultation on its plans to widen the use of biometric scanning systems. The minister for police the relevant minister has described the tool as the “most significant advance since genetic fingerprinting”.

Expert and Oversight Concerns

Abimbola Johnson, chair of the independent scrutiny and oversight board for the police race action plan, said: “There was very little discussion through equality strategy sessions of the facial recognition rollout despite clear relevance with the plan’s concerns.

“This disclosure demonstrate yet again that the anti-racism commitments the police has made via the equality initiative are failing to be integrated into wider practice. Our reports have warned that innovative tools are being rolled out in a context where ethnic inequalities, inadequate oversight and faulty information gathering continue to exist.

“Any use of this technology must adhere to rigorous official guidelines, be independently scrutinised, and prove it diminishes rather than compounds ethnic bias.”

Home Office Response

A government representative stated: “We takes the findings of the study seriously and we have already taken action. A updated software has been independently tested and procured, which has no statistically significant bias. It will be trialled early next year and will be subject to further assessment.

“Our priority is protecting the public. This gamechanging technology will assist officers to apprehend and prosecute offenders. There is officer review in every step of the procedure and no arrest or charge would be pursued without specialist personnel meticulously examining the results.”