Tech Beetle briefing GB

UK Police Forces Lobby to Use Biased Facial Recognition Technology Despite Known Issues

Essential brief

Key facts

UK police forces lobbied to continue using a facial recognition system known to be biased against women, young people, and ethnic minorities.

An NPL review found the system misidentifies Black and Asian individuals and women at significantly higher rates than white men.

Raising the confidence threshold reduced bias but also significantly decreased the number of investigative leads, leading to a reversal of that decision.

Critics warn that prioritizing operational convenience risks overriding fundamental rights and perpetuating racial disparities.

The Home Office is testing a new algorithm with no statistically significant bias and is consulting on expanding facial recognition use.

Highlights

UK police forces lobbied to continue using a facial recognition system known to be biased against women, young people, and ethnic minorities.

An NPL review found the system misidentifies Black and Asian individuals and women at significantly higher rates than white men.

Raising the confidence threshold reduced bias but also significantly decreased the number of investigative leads, leading to a reversal of that decision.

Critics warn that prioritizing operational convenience risks overriding fundamental rights and perpetuating racial disparities.

UK police forces have actively lobbied to continue using a facial recognition system despite evidence that it is biased against women, young people, and ethnic minorities.

The system, employed through the Police National Database (PND), compares a suspect's image against over 19 million custody photos to identify potential matches.

A Home Office-commissioned review by the National Physical Laboratory (NPL) revealed that the technology misidentifies Black and Asian individuals and women at significantly higher rates than white men.

In response, the National Police Chiefs’ Council (NPCC) initially raised the confidence threshold for matches to reduce bias, which decreased potential matches from 56% to 14%.

However, police forces complained that this adjustment led to fewer investigative leads, resulting in the reversal of the decision and a return to the original threshold despite the known bias.

The NPL study highlighted that false positives for Black women could be nearly 100 times more frequent than for white women under certain settings.

Critics, including former independent reviewers and race action plan leaders, have expressed concern that operational convenience is being prioritized over fundamental rights and that anti-racism commitments are not being effectively implemented.

The NPCC maintains that safeguards and training have been enhanced to mitigate bias and that the decision balances public protection with operational effectiveness.

The Home Office has stated that a new, less biased algorithm has been procured and will undergo testing and evaluation.

Meanwhile, a ten-week consultation is underway to consider expanding the use of facial recognition technology.

This controversy underscores the challenges of deploying facial recognition systems that must balance accuracy, fairness, and civil liberties in policing.