UK Home Office Backs AI Age Checks That Misclassify 2 in 3 16-Year-Olds
Updated
Updated · lighthousereports.com · Jun 19
UK Home Office Backs AI Age Checks That Misclassify 2 in 3 16-Year-Olds
3 articles · Updated · lighthousereports.com · Jun 19
Summary
More than two-thirds of 16-year-olds were classified as adults by Cognitec’s facial age tool in NIST benchmark data, according to an investigation into the UK Home Office system for asylum seekers.
The Home Office signed Cognitec last month and plans border trials next year after calling AI the most cost-effective way to challenge small-boat arrivals it says falsely claim to be children.
Leaked internal testing found even the best-performing vendor tended to overestimate teenagers’ ages, with 17-year-olds on average judged 18 or older, while female faces and Sub-Saharan Africans saw worse results.
Sub-Saharan girls showed average errors of 4.6 years in the Home Office report, raising the risk that a 14-year-old could be treated as an adult and placed in adult accommodation or detention.
The findings come as the UK’s advisory age-assessment committee was disbanded before the AI plan was announced, and rights groups warn the policy could become a model for wider European use.
With known racial biases, how will the UK's new AI border guard protect children of colour?
If experts call it 'hideously inaccurate', why is the UK using AI to judge child asylum seekers?
UK to Deploy AI Facial Age Estimation at Borders in 2027: Technology, Bias, and Safeguards
Overview
The UK plans to introduce AI-driven Facial Age Estimation (FAE) at its borders in 2027 to help staff identify adults posing as children among asylum seekers. The system analyzes photos using AI trained on millions of images to spot age-related facial patterns, aiming to replace invasive scans and speed up assessments. However, FAE is not fully reliable—it has a margin of error, especially for 16- to 18-year-olds, and can be biased against certain groups, like Sub-Saharan African girls. Because of these risks, final decisions will remain with trained human officials, who must apply the benefit of the doubt in complex cases.