UK Plans to Deploy Flawed Facial Age Estimation on Asylum Seekers Despite Known Bias

UK Deploys Flawed Facial Age AI Despite Known Accuracy Crisis

The United Kingdom Home Office is moving forward with facial age estimation technology at its asylum borders beginning in 2027, despite internal government testing documenting significant accuracy failures—particularly against Sub-Saharan African minors. According to Wired and Lighthouse Reports’ investigation, the system overestimated the age of female Sub-Saharan African children by an average of 4.6 years, meaning a 13.5-year-old girl could be classified as an 18-year-old adult. Misclassification carries life-altering consequences: children wrongly identified as adults lose statutory protections and face placement in adult-only detention centers.

The Leaked Testing Report and Its Findings

Lighthouse Reports obtained an internal Home Office document detailing trials of seven facial age estimation algorithms conducted in 2025. The report examined the “best performing” of these systems, though it withheld the companies’ identities. According to Wired’s reporting, the testing revealed dramatic performance disparities across demographic groups. The Sub-Saharan African population bore the largest margin of error—a pattern with direct operational impact, since Home Office data cited in the investigation shows Sub-Saharan Africans represent the majority of asylum seekers requiring age assessments in 2025, particularly those arriving via small boats across the English Channel.

The accuracy issues were not evenly distributed. Female Sub-Saharan African applicants experienced the worst outcomes, with predictions off by 4.6 years on average. Male assessments and performance on other demographic cohorts showed markedly better results, indicating the system encodes systematic bias rather than random noise.

Institutional Dismissal of Warnings

The Home Office response to these findings underscores a pattern of institutional risk acceptance. According to the investigation, the department disbanded its scientific advisory committee—the body tasked with evaluating the technology’s readiness and safety—and proceeded without apparent course correction. This structural removal of internal oversight came despite the leaked report’s evidence that high-stakes predictions would fall on the population group with the worst system performance.

Wired notes that the deployment occurs amid a global wave of anti-migrant policy and surveillance spending, particularly under the second Trump administration. Governments worldwide are accelerating the adoption of automated decision systems applied to vulnerable populations—often without those populations’ knowledge of the technology’s operation or appeal mechanisms.

Why This Matters

The UK’s deployment signals a critical policy inflection: governments are moving facial age estimation from digital gatekeeping (age verification for websites) into physical infrastructure with custodial stakes. An error in this context is not a failed login attempt; it is removal of a child’s legal status and institutionalization in adult facilities. The 4.6-year margin of error on the population representing the majority of assessments means the system is not a safety supplement to human judgment—it is a decision mechanism that systematically ages out minors in a specific demographic group. If independent reproduction confirms Wired’s findings, this deployment represents a documented case of knowingly advancing biased automated decision-making into life-altering immigration proceedings, despite internal evidence that the bias falls hardest on the least-resourced asylum seekers.