UK plan to use AI to judge migrants' ages puts child refugees at risk, charities warn

Key points

• The British Home Office plans to introduce AI facial age-estimation software to screen migrants arriving in the UK.
• Rights groups and children's charities warn the technology carries racial bias that systematically overestimates the ages of Black children.
• Critics say the error will place children in adult facilities, exposing them to serious harm.
• No peer-reviewed trial data for this specific deployment has been published; the warnings rest on known flaws in comparable systems.

A child fleeing a war zone arrives in the UK alone. An algorithm looks at their face and decides they are an adult. They are sent to live with grown strangers in accommodation not designed to protect them. That is the scenario rights groups are now warning is likely if the Home Office proceeds with its plan to use AI-powered facial age-estimation, a technology that reads a person's face and tries to calculate how old they are, to screen migrants.

The Human Rights Network and several children's charities, as first reported by The Guardian, are urging ministers to drop the plan. Their core argument: the AI models behind these tools are trained mainly on faces from wealthier countries, which means they perform less accurately on darker skin tones and systematically read Black faces as older than they are.

What is the actual risk to children?

Children wrongly assessed as adults lose the legal protections the UK gives unaccompanied minors. They can be housed with adults, denied specialist support, and processed through the asylum system without a guardian. For a child already traumatised by conflict, the consequences can be severe.

Age-estimation technology is genuinely difficult even for experienced clinicians. Human assessors already make errors. Adding an AI layer that carries documented racial bias, critics say, does not make the process more accurate. It makes a biased outcome faster and harder to challenge.

What does the evidence say about bias in these systems?

Researchers have documented for years that facial-analysis AI performs unequally across skin tones. A widely cited 2018 study by MIT researcher Joy Buolamwini found that commercial face-analysis tools misclassified darker-skinned women at error rates up to 34 percentage points higher than for lighter-skinned men. Age-estimation tools, which are more complex than simple recognition, carry similar risks.

The Home Office has not published independent, peer-reviewed trial results showing its chosen system performs equally across ethnicities. That absence matters enormously when the stakes are a child's safety.

What happens next?

The government has not reversed course. Rights groups are calling for the plans to be scrapped pending transparent, independent auditing of any tool's accuracy across all skin tones and age groups, with children specifically included in that testing.

For anyone working with or advocating for unaccompanied child migrants, the practical message is clear: challenge any age assessment that relied on automated facial analysis, ask what system was used and whether it has published bias audits, and insist on human review by a qualified assessor.

The technology may eventually improve. Right now, the evidence that it works fairly for the children most likely to face it simply is not there.

Common questions

Can facial AI reliably tell how old someone is?

Not with the accuracy needed for a legal decision. Even leading commercial systems carry measurable error rates, and those errors are not distributed equally across ethnicities.

Are there safeguards in place?

The Home Office has not published details of independent bias auditing for its planned system. Rights groups say that audit must happen before any deployment, not after.