UK deploys facial age estimation AI for asylum seekers despite known accuracy flaws
The UK Home Office announced plans to use facial age estimation AI to check asylum seekers' ages at the border from 2027, despite an internal April 2025 report showing the technology significantly misclassifies children as adults, particularly those from Sub-Saharan Africa. The leaked assessment found error rates averaging 4.6 years for female migrants from the region, potentially classifying 13-14 year-olds as adults, and the government proceeded despite scientific advisers expressing concerns about accuracy and political rather than evidence-based decision-making.
What happened
The Home Office announced in May 2026 that facial age estimation AI would be deployed at UK borders from 2027 to check the ages of asylum seekers arriving by small boat.
An internal Home Office report from April 2025, obtained through investigation, revealed the technology was significantly flawed. The report showed that for female Sub-Saharan African migrants, the AI's age predictions were off by an average of 4.6 years, meaning a 13-14 year-old could be assessed as an adult.
Testing by Lighthouse Reports found the technology misclassified more than a third of 16-year-olds as adults in some cases, with error rates reaching 70 per cent in certain tests. The report also found the system performed substantially worse for Sub-Saharan Africans, the largest group of migrants subject to age assessment in 2025.
Reported impact
Asylum seekers incorrectly classified as adults by the AI could lose legal protections and be placed in adult-only detention centres. The technology appears to contain systematic bias affecting vulnerable migrant children from Sub-Saharan Africa, regions with the highest numbers of small boat arrivals.
Scientific advisers to the Home Office, including Professor Tim Cole at University College London's Institute of Child Health, expressed concerns that the government is deploying the technology for political rather than evidence-based reasons despite being aware it is 'hideously inaccurate'.
Index analysis
The AI system exhibits clear racial bias, with substantially higher error rates for Sub-Saharan African migrants compared to other groups. The technology also appears less accurate for individuals showing visible signs of stress from travel.
The Home Office acknowledged the technology's limitations in its April 2025 internal assessment but proceeded with deployment plans, describing the same system as 'cost-effective' with 'promising performance and accuracy' when announcing the scheme publicly.
Governance controls
Scientific advisers to the Home Office were not consulted on the deployment decision, reportedly to avoid criticism of the technology. The government obtained a leaked internal report in April 2025 that detailed the system's significant flaws but proceeded with the announcement of deployment from 2027.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.