UK Home Office to use AI age estimation on asylum seekers – how accurate is the technology?

The UK Home Office announced plans to use AI-powered facial age estimation for asylum seekers from 2025, marking a high-stakes application of the technology. The system, developed by Cognitec via Akhter Computers, analyzes facial features like skin texture and bone structure to estimate age ranges, producing probabilities rather than definitive answers. Trials are set to begin at a Dover processing facility before wider rollout, with officers retaining final decisions. The technology relies on algorithms trained on millions of images, achieving a mean absolute error of under three years—better than human accuracy, which averages eight years. However, National Institute of Standards and Technology (Nist) data shows errors spike near the 16-to-18 threshold, the exact boundary determining child protections under UK law. Female faces also face consistent accuracy gaps, raising concerns about fairness. Under current UK law, unaccompanied asylum seekers under 18 are classified as children, receiving care, education, and legal safeguards. Misclassification risks depriving vulnerable individuals of protections or exposing minors to adult systems. The Home Office emphasizes the AI will be one factor among others, but research on automation bias suggests officers may over-rely on algorithmic suggestions under pressure. Cognitec ranks fourth globally in Nist’s facial age estimation benchmarks, with Regula topping recent tests. The UK’s system will use border-crossing photographs, a dataset Nist has evaluated since 2024. While the technology improves over time, its limitations near critical age boundaries remain a point of contention among advocates and policymakers.