UK Home Office to Trial AI Facial Age Estimation on Asylum Seekers Despite Significant Accuracy Concerns
Key Takeaways
- ▸Yoti's facial age estimation model has a mean absolute error of 1.88 years for teens, with individual predictions potentially off by 5-6 years—enough to determine placement of asylum seekers
- ▸Error distributions are asymmetric and wide, meaning average accuracy metrics mask serious real-world disparities that disproportionately affect vulnerable populations
- ▸The trial proceeds despite legal opinions questioning the lawfulness of existing Home Office AI systems in asylum processing
Summary
The UK Home Office announced on April 28, 2026, that it will proceed with a trial of artificial intelligence facial age estimation technology to assess the age of migrants arriving via the Channel. The system, developed by British identity firm Yoti, will use deep neural networks to estimate age from facial photographs, potentially determining whether asylum seekers are placed in children's facilities or adult accommodations.
However, the announcement has triggered widespread alarm from human rights organizations and legal experts. Independent testing shows Yoti's model has a mean absolute error of approximately 1.88 years for ages 13-16, meaning predictions can vary significantly from actual age in real-world distributions. The article highlights that this metric obscures the true risk: error distributions are wide and asymmetric, with some predictions potentially off by five or six years—a difference that fundamentally alters a person's legal status and access to protections.
The trial arrives amid damning reports about the Home Office's historical failures in age assessment and emerging legal opinions suggesting existing AI deployment in asylum processing may already violate UK law. Human Rights Watch and migrant rights charities have condemned the plan, arguing it constitutes an "AI experiment on children seeking asylum." Legal experts question whether current accuracy standards meet the threshold required for deployment in a system affecting vulnerable minors who have no means of formal appeal or legal representation.
- Human rights organizations warn the deployment lacks adequate transparency, accountability, and appeal mechanisms required for life-altering decisions on children
- The Home Office's decade-long history of poor age assessments provides context for concern that AI automation will entrench rather than remedy systemic failures
