← Back to feed

The UK Home Office has awarded a contract to develop and test AI-powered facial age estimation for use at the border, with deployment planned for 2027. The technology will analyse photographs taken at border entry to estimate whether migrants claiming to be children are adults.

The Numbers Behind the Decision

The Home Office is acting on a documented problem, not a hypothetical one:

• In the year ending March 2026, more than 6,400 migrants claiming to be children were age assessed at the UK border
• 43% were found to be adults after assessment
• In the year ending June 2025, 111,084 people claimed asylum in the UK — up 14% year on year

The stakes are material. Unaccompanied child migrants are housed in the local council care system and receive legal protections that simplify the asylum application process and improve residency outcomes. Adults placed in the care system displace children who are entitled to that support.

A 2025 report by the UK’s independent immigration inspector found the assessment process was failing in both directions: some adults were classified as children, and some children were incorrectly classified as adults. The report acknowledged that without a “foolproof” test, some errors were “inevitable.”

What the Technology Does

The system takes photographs of migrants at the border and runs facial analysis to estimate age. The contract was awarded to an unnamed software company for development and testing. The Home Office described initial testing as showing “promising performance and accuracy.”

No accuracy figures have been published. No false-positive or false-negative rates have been disclosed. The system is in development, not deployed.

Human Rights Watch called for the scheme to be scrapped, describing it as “unproven technology” that will undermine protections for vulnerable children. Critics argue that facial analysis for age estimation is unreliable across different ethnicities and can be systematically biased.

The Policy Logic

The UK government’s argument is operational: the manual assessment process is inconsistent, resource-intensive, and still produces significant error rates. An AI pre-screening system would not replace human decision-making but would flag cases for more detailed assessment.

The deployment sits within a broader European pattern. AI is increasingly being applied to asylum and immigration processing — document verification, biometric registration, risk scoring. Each case presents the same tension: AI can process higher volumes with greater consistency, but errors at this scale affect legal rights, not just efficiency metrics.

In the UK context, the error cost is asymmetric. Misclassifying an adult as a child creates a resource and legal processing problem. Misclassifying a child as an adult means a minor loses access to care system protections and is processed through the adult asylum track. The latter is the harder error to quantify but the more consequential one.

A 2026 deployment means the system will operate under the UK’s current AI regulatory framework, which has no statutory pre-deployment evaluation requirement for government AI tools in immigration. The European Union AI Act, which classifies biometric categorisation of individuals as high-risk, does not apply to UK border systems.

The Home Office expects to have more to say on evaluation methodology before the 2027 launch date.