Report warns of risks in using AI facial age estimation for child asylum seekers

A coalition of refugee and children's rights organisations has warned that the Government's proposed use of artificial intelligence (AI) to estimate the age of asylum seekers could create new risks for children unless accompanied by robust safeguards.

The 11-page Refugee and Migrant Children's Consortium (RMCC) report, Benchmarks and Borders: The use of Facial Age Estimation to assess the age of unaccompanied young people seeking asylum, can be downloaded here.

As we reported on EIN earlier this month, the Home Office announced in May that it had awarded a contract for Facial Age Estimation (FAE) technology, with further trials planned during 2026 ahead of a potential rollout in 2027. According to the Government, the technology will be used as an additional source of information for immigration officers rather than replacing human decision-making.

The RMCC's new report emphasises that determining whether a young asylum seeker is a child or an adult is "of fundamental importance to their life," affecting access to accommodation, education, safeguarding arrangements and the processing of asylum claims.

While acknowledging that AI-based age estimation could potentially assist initial decision-making, the report warns that significant concerns remain about accuracy, bias, transparency and legality.

The report highlights concerns about the accuracy of facial age estimation for older teenagers, particularly those aged 16 and 17, who make up the majority of unaccompanied child asylum applicants in the UK. It notes that age estimation remains difficult and that "absolute errors of several years remain common," particularly around the age of 18, where decisions can determine whether an individual is treated as a child or an adult.

In addition, existing AI systems are unlikely to have been tested on populations comparable to unaccompanied asylum-seeking children, with the report stressing: "FAE is inherently complex because of the significant variations in aging characteristics across ethnicities, genders and lifestyles. … [M]any of these young people may have experienced undernutrition, illness, chronic stress, trauma, and long and dangerous journeys. These experiences can affect growth, appearance and visible development. In the late teenage years, where the physical differences between a child aged 16 or 17 and a young adult aged 18 or 19 may already be very subtle, this creates a particular risk of error."

The report further raises concerns about potential demographic biases, image quality issues and the possibility that immigration officials could place excessive weight on algorithmic outputs, a phenomenon known as "automation bias."

The RMCC also points to existing problems in age assessment decisions. Citing Home Office statistics, it notes that between July and December 2025, at least 326 children were initially assessed as adults before later assessments concluded they were under 18.

The report calls for any use of FAE to remain advisory rather than determinative, and recommends that children continue to receive the benefit of the doubt where uncertainty exists. It also calls for independent validation, data protection assessments, child rights impact assessments and transparent testing before any operational deployment.

In its conclusion, the report states: "If FAE is used at all, it must remain only one part of a wider, child-centred process. It should never be used as the sole or determinative basis for deciding that a young person is an adult, nor should it replace the crucial role of social workers in assessing age in a holistic manner. This is particularly important for young people aged 16 or 17, where the consequences of error are especially serious."

Separately, an investigation published last week by Lighthouse Reports and co-published with WIRED and The Independent warns that FAE is "deeply unreliable".

Drawing on a leaked Home Office evaluation and independent testing conducted by the US National Institute of Standards and Technology (NIST), the investigation reported significant error rates in age estimation systems, particularly for teenagers close to the age of 18.

According to the investigation, Cognitec Systems, the company contracted by the Home Office to supply the technology, misclassified more than two-thirds of 16-year-olds as adults when tested using photographs from the US border. The analysis also found that performance varied across demographic groups, with higher error rates reported for people from African countries.

The investigation further reported that an internal Home Office assessment found that the best-performing system tended to overestimate the ages of teenagers, with 17-year-olds on average being assessed as 18 or older. It also identified higher error rates for female faces and for people from Sub-Saharan Africa, with average error rates for Sub-Saharan African girls of 4.6 years.

The Independent reported last week that Professor Tim Cole, emeritus professor of medical statistics at University College London's Institute of Child Health and a former member of the Home Office's Age Estimation Science Advisory Committee (AESAC), questioned both the reliability of FAE and the Government's decision to disband the committee shortly before ministers announced plans to use AI-assisted age assessments.

Cole told The Independent that committee members had wanted to raise concerns about the limitations of FAE but were not given the opportunity to do so before the advisory body was dissolved. He said he suspected the committee had been stood down because its members "would have been quite damning about" the technology.

Cole was quoted as saying: "We were keen to highlight the inadequacies of facial age estimation, but this opportunity was not presented to us, and then the committee was shut down, I suspect for the reason that we would have been quite damning about it. Their argument was that our expertise was very specialised and not relevant to facial age estimation. In reality, what facial age estimation is doing is trying to interpret the age-related changes in appearance, which are fundamentally biological, and our background is biological growth and ageing."

Human Rights Watch's senior technology and human rights researcher, Anna Bacciarelli, said the use of FAE was too risky to pursue and would severely endanger the human rights of children seeking asylum.

Calling on the Home Office to scrap plans for its use, Bacciarelli stated: "To use this for life-changing decisions in refugee processing centres is to introduce an unreliable, untested technology into an already flawed process. .. This technology has no place in deciding whether a young person can access the rights and protections they are entitled to."