Rights group warns against UK facial age estimation plans for asylum process – JURIST

Human Rights Watch (HRW) urged the UK Home Office on Friday to abandon its 2027 plans to use facial age estimation (FAE) AI to assess young asylum seekers, warning that the technology risks discriminatory errors and sets a dangerous global precedent for migration policy.

In an open letter to the UK Home Office, HRW, Foxglove, and 61 civil society groups raised substantial concerns about the discrimination in FAE, particularly in relation to women and people of color. The FAE is a nascent technology that UK shops and bars currently deploy to prevent minors from accessing age-restricted products, like cigarettes and alcohol. The rights group stated, “To use this for life-changing decisions in refugee processing centers is to introduce an unreliable, untested technology into an already flawed process.”

The UK Home Office stressed that the use of FAE does not automate or replace what remains an integral human decision. Immigration officers remain responsible for considering several factors, such as a person’s demeanour as well as appearance. Immigration officers apply a “benefit of the doubt” approach in age disputes. They treat a person as an adult only when two officers independently conclude that the person’s appearance and behavior clearly show they are well over 18. If they cannot reach that threshold, they treat the person as a child and refer them to a local authority for a more detailed, Merton-compliant age assessment.

Nevertheless, Foxglove Co-Executive Director Martha Dark accused the government of recognizing the inaccuracy of its AI tools, stating that the government’s own internal documents reveal that the tools frequently misidentify children as adults. She further alleged that the systems are racially biased, showing particular accuracy issues when tested on individuals from countries in Sub-Saharan Africa. Dark emphasized, “Children seeking asylum have often suffered unimaginable trauma. They should not be the test subjects for experimental tech that has baked-in inaccuracy and racist bias.”

A study from the National Institute of Standards and Technology (NIST) observed that age-estimation algorithms vary in accuracy depending on factors such as image quality and demographic characteristics, with performance differing across groups. It also reported consistently higher error rates for female faces, but the underlying reasons are unknown.