Flawed AI Scans Border Seekers While Agent Servers Fall

Here is the part that should stop you cold: the British government's own internal testing found that its facial age estimation system guesses the age of young Sub-Saharan African girls wrong by an average of 4.6 years. That means a 13-year-old could be logged as an 18-year-old adult. And the government is pushing ahead anyway.

Starting next year, the UK plans to use AI-powered facial scanning to help determine the ages of asylum seekers arriving without identity documents. It is believed to be the first time any government has deployed this kind of technology specifically for immigration age assessments. The stakes are not abstract — children who are incorrectly classified as adults can lose key legal protections and be placed in adult detention facilities.

A joint investigation by WIRED, Lighthouse Reports, and The Independent obtained a leaked Home Office document detailing tests of seven different facial age estimation algorithms. The report focused on the best-performing system of the seven, and even that one showed significant accuracy problems and troubling bias. The technology consistently performed worse when assessing Sub-Saharan African individuals compared to other groups.

That bias is not a minor footnote. Sub-Saharan Africans represent the largest group of migrants crossing the English Channel into the UK in small boats, and they had more age assessments raised against them in 2025 than any other regional cohort, according to Home Office data. In other words, the group most likely to be scanned is also the group the technology works worst on.

What makes this even harder to square is that the Home Office apparently knew what it was getting into. According to the investigation, the department dissolved its own scientific advisory committee — the one tasked with providing independent guidance on age estimation methods — while it was actively exploring AI adoption. That is not an oversight. That is a choice.

The broader context here matters too. Governments worldwide are pouring money into surveillance technology and deploying it against migrant populations who often have no idea the tools are being used, how the systems work, or how to challenge a result they believe is wrong. When the technology has documented accuracy gaps and the oversight bodies have been quietly shut down, that is not a border security upgrade. That is a due process problem wearing a tech hat.

Age verification AI has been spreading fast — social media platforms, adult content sites, and now physical borders. Each new deployment carries its own risk profile. But few contexts are higher stakes than determining whether a frightened teenager gets treated as a child or an adult by the state. Getting that wrong does not generate a help ticket. It changes a life.