Benchmarks and Borders: The use of Facial Age Estimation to assess the age of unaccompanied young people seeking asylum

Determining whether a young person seeking asylum is a child or an adult is of fundamental importance to their life: it determines whether they receive care from children’s services; can access suitable accommodation and education; safeguarding arrangements; and the way their asylum claim is processed. Errors at the border, where the age of unaccompanied children and young people is too often decided on the basis of ‘appearance and demeanour’, have repeatedly resulted in children being treated as adults, with serious safeguarding consequences. The Independent Chief Inspector of Borders and Immigration (ICIBI) has criticised these visual assessments as perfunctory and lacking cultural awareness and consistency, and the evidence from NGOs has consistently shown that hundreds of children each year are wrongly placed in adult settings, including immigration removal centres and adult prisons. Any reform of age assessment processes must reduce, not reproduce, those harms.

Age assessments have long been an area of contention, and there is no ‘quick fix’ for a process that can be complicated and lengthy. After many years of exploring the use of highly criticised ‘scientific methods’ of assessing age, such as X-rays, in July 2025 the government announced proposals to use artificial intelligence (AI) to assess age instead, as “the most cost-effective option”. On 29^th May 2026, the government announced that the procurement process for the Facial Age Estimation (FAE) algorithm had concluded and a contract awarded to Akhter Computers Ltd. FAE will be implemented as part of initial age decisions at the border with “further trials and testing throughout 2026” before a planned roll out in 2027.

The Home Office has made clear that the technology will be an additional source of information for immigration officers rather than a replacement for human decision-making or social work age assessments. However, the information shared to date does not provide the level of detail needed to assess whether FAE can be used safely, fairly or lawfully in this context, including the testing methodology, datasets used, performance across demographic groups, false-adult rates for 16- and 17-year-olds, or the safeguards that would apply before any operational rollout.

The development of FAE also sits within a wider government agenda to “strengthen” age assessments and “root out” allegedly false age claims, including proposals expected to be taken forward through forthcoming immigration and asylum legislation. This makes it particularly important that any technological or legislative changes are assessed against the evidence of existing safeguarding failures, and that safeguards for children are strengthened rather than weakened.

In light of extensive evidence showing how flawed human ‘visual assessments’ at the border are at present, the use of AI for age estimation in the UK asylum process could, in theory, benefit the triage process if used narrowly and with robust safeguards. However, given the lack of accuracy with existing systems, its use must be approached with caution. The use of AI also introduces new technical and rights risks that must be understood and addressed. This briefing aims to summarise the existing evidence and propose safeguards that must be put into place if FAE is ever to be used.

Key findings:

While errors have improved over a decade, estimating age using FAE remains subject to significant margins of error. Young people aged around 18 remain some of the hardest cases, and false-adult risks vary by the threshold chosen (e.g., challenge-25/28). The technology also performs less well when images are taken in real-world conditions, such as at borders where lighting, camera angle and facial position can vary, compared with more controlled settings like office-style photos. FAE must also take into account the traumatic journeys that young people will have experienced on their journeys to the UK. These assessments often happen through the night and exhaustion and trauma are known to “age” children. Accuracy can also differ between groups, including by sex and ethnicity.
Because of the risks that FAE can wrongly identify a child to be an adult, just as human age estimation has for a number of years, there are a number of procedural safeguards that must remain if FAE is to be used, including giving the benefit of the doubt; referral to social workers for more detailed assessment; access to an appropriate adult and legal representation; and the ability to legally challenge any decision. Any use of AI must sit within, not in place of, a lawful and child-centred process.
If FAE is considered at all, it should be advisory only, used with conservative thresholds so that a child is only at risk of being treated as an adult in tightly defined, case-specific circumstances, where the evidence clearly and consistently indicates that they are significantly over 18 and there is no credible evidence to the contrary. FAE should never be solely used to determine age.
Use of technology must be subject to independent validation on UK relevant cohorts and border quality imagery, and embedded within a transparent, rights compliant procedure.

Read the full briefing here.