The UK Will Scan Asylum-Seekers’ Faces for Age Checks—Despite Knowing the Tech Is Flawed

The UK Home Office intends to deploy facial age estimation AI at the border to assess asylum seekers, despite internal tests showing the technology regularly misclassifies children as adults and performs worst on the largest migrant group, according to documents obtained by WIRED and Lighthouse Reports [1]. The internal report, produced in April 2025, tested seven algorithms against more than 2.5 million images [1]. It found the "best performing algorithm" had "substantial deviations" when tested on images of Sub-Saharan Africans, the largest cohort subject to age assessments in 2025 [1]. For female Sub-Saharan Africans, the system's age estimate was off by an average of 4.6 years, meaning a 13.5-year-old girl could be assessed as an 18-year-old adult [1]. The report also noted that on average, the system tended to predict a 17-year-old would be over 18 [1]. The findings raise questions about the technology's reliability in high-stakes scenarios. The Home Office disbanded a scientific committee designed to advise on age estimation methods while it was exploring AI introduction [1]. "We were keen to highlight the inadequacies of facial age estimation, but this opportunity was not presented to us, and then the committee was shut down," said Tim Cole, an emeritus professor of medical statistics at University College London's Institute of Child Health and former committee member [1]. Cole described the face scans as "hideously inaccurate" [1]. Since 2010, 40 percent of people who faced age assessments were classed as adults, according to official statistics [1]. The leaked report indicates that testing primarily used high-quality images, and accuracy rates would likely be worse in practice, as photos taken at initial encounters were "routinely worse" than follow-up images [1]. The report was unable to determine whether photo quality or the physical condition of asylum seekers at arrival had more impact on results [1]. Privacy and AI safety concerns extend beyond this deployment. Research on AI model documentation reveals systematic gaps in safety-critical disclosures, with categories such as deception behaviors, hallucinations, and child safety evaluations accounting for 148, 124, and 116 aggregate points lost respectively across evaluated models [3]. The broader context of internet privacy underscores that personally identifiable information, including age and physical address, can uniquely identify individuals even without explicit names [2]. "Children seeking asylum have often suffered unimaginable trauma," said Martha Dark, co-executive director of rights group Foxglove [1]. "They should not be the test subjects for experimental tech that has baked-in inaccuracy and racist bias" [1]. Foxglove and 61 other organizations sent an open letter to the government asking the Home Office to scrap the plan [1]. A Home Office spokesperson said the department has "rigorous processes in place to verify an individual's age" and is working to modernize them through testing of "fast and effective facial age estimation technology" [1]. The spokesperson added that the committee was disbanded due to requiring "different fields of expertise" [1]. The government has delayed rollout until 2027 [1].