TL;DR
Children’s charities and rights groups have urged ministers to abandon plans to screen migrants with AI facial age-estimation, warning that the underlying models are biased and will push black children arriving from conflict zones into the adult system. The government has conceded that even the best systems carry a margin of error of up to 30 months.
The problem predates the technology. Helen Bamber Foundation figures released in May showed 755 children were incorrectly identified as adults on arrival in the UK during 2025, based on Home Office data. Maddie Harris of the Humans for Rights Network, whose charity has worked with hundreds of children wrongly routed into the adult system by immigration officers, said automation would not improve those numbers. “This is not about protecting children – this is about shoring up the decisions that [the authorities] are making,” she said, calling it a cheap process.
The consequences of a wrong classification are immediate. A child treated as an adult becomes exposed to detention and removal, and unaccompanied children would be placed in accommodation alongside adults at a time when asylum accommodation is being targeted by protests.
The contract announced in May uses technology from Cognitec, a German firm. A Lighthouse Reports investigation in June found the system showed bias against sub-Saharan Africans, tending to overpredict the ages of minors — which matters most for children from countries such as Sudan and Somalia. The tech monitor Foxglove said the systems rely on racist bias and that children should not be used to test experimental technology. The Refugee and Migrant Children’s Consortium noted that trauma can age a child’s appearance.
Petra Molnar of Migration and Tech Monitor placed it in a longer sequence of European failures, recalling AI lie-detection pilots for refugee screening that proved to be “snake oil”. She compared facial age estimation to phrenology: “It’s effectively ‘neo-phrenology’ by categorising people based on physical features.”
The Home Office said age assessments are a vital border security tool and that it is exploring the technology to make the process more consistent.
Looking Forward
An admitted 30-month error band is the number to hold onto, because it is wide enough to place a 15-year-old on either side of the adult threshold. The system runs alongside immigration officer checks rather than replacing them, which means the practical question is whether officers treat a machine estimate as corroboration of a judgement they had already formed.