Don't replace human error with machine error
In the long and contested history of how nations receive those who flee, the United Kingdom now proposes to delegate one of its most consequential judgments — whether a frightened young person is a child or an adult — to an algorithm reading a face. The Home Office has awarded a £322,000 contract to develop AI age-estimation technology for asylum seekers arriving at Dover, with a 2027 rollout planned, even as more than a hundred refugee and children's rights organisations warn that the system risks automating and amplifying a mistake the border already makes too often. At stake is not efficiency, but the fate of children who may be wrongly placed among adults in detention — a harm that, once done, is rarely easily undone.
- Over 100 refugee and children's rights organisations have raised urgent alarms that AI facial analysis cannot account for the physical toll — trauma, malnutrition, exhaustion — that makes young asylum seekers appear older than they are.
- The government's own data already shows that hundreds of children are wrongly classified as adults through existing visual assessments at the border, with devastating consequences for their safety.
- The Home Office frames the technology as a tool to catch adults falsely claiming to be minors, insisting final decisions will remain with immigration officers and that rigorous testing will precede any rollout.
- Charities are not calling for an outright ban but demanding strict safeguards — AI in an advisory role only, access to legal representation, and the right to challenge decisions made with algorithmic input.
- Critics warn the technology risks replacing human error with machine error at scale, creating a false sense of certainty in decisions that are already among the hardest — and highest-stakes — the system faces.
The Home Office has awarded a £322,000 contract to develop artificial intelligence capable of estimating the age of young asylum seekers from facial photographs, with a national rollout planned for 2027. The system would process images already taken of small-boat arrivals at Dover, producing age estimates in seconds. More than a hundred refugee and children's rights organisations have responded with alarm.
Their concern is grounded in existing evidence. Home Office data shows that young asylum seekers assessed by social workers are recorded as children at more than twice the rate of those assessed by immigration officers at the border. The Refugee and Migrant Children's Consortium warns that AI cannot account for the physical effects of the journeys these young people have survived — the trauma, malnutrition, and exhaustion that can make a teenager appear years older. More than two-thirds of lone child asylum seekers arriving in the UK are aged 16 or 17, a group whose age is already genuinely difficult to judge.
The government argues the technology will identify adults falsely claiming to be children in order to access protections meant for minors. Border Security and Asylum Minister Alex Norris said it would ensure those "gaming the system" are detained and removed, while genuine cases receive protection. The Home Office insists the algorithm will not make final determinations — immigration officers will retain that authority — and that testing will precede any deployment.
The consortium's report, "Benchmarks and Borders," does not oppose AI outright but insists it must function only in an advisory capacity, never replacing comprehensive assessments by trained social workers. It calls for legal representation, the presence of an appropriate adult, and the right to challenge decisions. Senior policy analyst Kama Petruczenko of the Refugee Council noted that the government's own figures already document hundreds of children wrongly classified as adults through flawed visual assessments — and warned the technology risks producing "a false sense of certainty in decisions that are already extremely difficult to get right."
The contract has been awarded to Akhter Computers Ltd for further development and testing. For the coalition of organisations watching closely, the deeper question is whether automation will correct a systemic failure or simply accelerate it — and whether the government will act on the warnings before vulnerable children pay the price.
The Home Office has awarded a contract worth £322,000 to deploy artificial intelligence technology that will attempt to determine the age of young asylum seekers by analyzing their facial photographs. The system, set to roll out in 2027, will process images already taken of small-boat arrivals at Dover, estimating age in seconds through algorithmic analysis. But more than a hundred refugee and children's rights organisations are sounding an alarm: the technology risks automating a mistake that the system already makes with troubling frequency—wrongly classifying children as adults, with consequences that can be severe and irreversible.
The concern is not abstract. Home Office data shows that young asylum seekers assessed by social workers are recorded as children at more than twice the rate of those assessed by immigration officers at the border. More than two-thirds of lone child asylum seekers arriving in the UK are aged 16 or 17, a cohort whose appearance can be deceptively difficult to judge. The Refugee and Migrant Children's Consortium, whose member organisations advocate for the rights of refugee and migrant children, released a report in June warning that AI cannot account for the physical toll of the journeys these young people have endured—the trauma, the malnutrition, the exhaustion that reshape how they look. To treat facial analysis as a reliable proxy for age is to ignore the lived reality of flight.
The government's framing emphasises a different concern: adults falsely claiming to be children in order to access protections and resources meant for minors. Border Security and Asylum Minister Alex Norris stated that the technology would identify those "gaming the system" and ensure they are "detained and removed without delay," while those who genuinely qualify for protection would receive it. The Home Office acknowledges the need to safeguard minors and says the technology will undergo testing before rollout. Final age determinations will still be made by immigration officers, not by the algorithm alone.
Yet the consortium's report, titled "Benchmarks and Borders," does not rule out AI use entirely. Instead, it urges caution: the technology should function in an advisory capacity only, never as a substitute for comprehensive age assessments conducted by trained social workers. The report calls for safeguards including access to legal representation, the presence of an appropriate adult, and the right to challenge decisions. Most fundamentally, it warns against replacing human error with machine error—a substitution that could leave more children wrongly placed in adult accommodation, detention centres, or prisons.
Kamena Dorling, co-chair of the consortium, described the government's proposals as "deeply concerning," noting that AI exhibits the same patterns of bias and inaccuracy as human decision-making. Kama Petruczenko, a senior policy analyst at the Refugee Council, pointed out that the government's own figures already document hundreds of children being wrongly classified as adults following flawed visual assessments at the border, with devastating consequences for their safety and wellbeing. Poor image quality and bias in the datasets used to train the AI could compound these errors. The technology, she warned, risks creating "a false sense of certainty in decisions that are already extremely difficult to get right."
The contract has been awarded to Akhter Computers Ltd, which will further test and develop the system before the planned 2027 rollout. The Home Office maintains that rigorous testing, evaluation, and assurance will precede any national deployment. But the coalition of organisations sees a troubling trajectory: a system already prone to misclassification being accelerated and scaled through automation, potentially ensnaring more vulnerable young people in the adult detention system. The question now is whether the government will heed the warnings or proceed with a technology that promises efficiency at the risk of deepening an existing harm.
Notable Quotes
AI cannot account for the factors that can significantly affect a young person's appearance after fleeing conflict and persecution, including trauma, malnutrition, and exhaustion.— Kamena Dorling, co-chair of the Refugee and Migrant Children's Consortium
There is a real danger that this technology creates a false sense of certainty in decisions that are already extremely difficult to get right. If flawed assessments are simply automated, more children could end up wrongly placed in adult accommodation, detention centres or even prisons.— Kama Petruczenko, senior policy analyst at the Refugee Council
The Hearth Conversation Another angle on the story
Why does age assessment matter so much in asylum cases? What actually happens if a young person is wrongly classified as an adult?
If you're classified as an adult, you lose access to protections designed for children—legal representation, social workers, safeguarding. You can be placed in adult detention facilities or even prisons. For a traumatised teenager who's already fled violence, that's not just bureaucratic error. It's a second violation.
The government says the AI is just advisory, that humans make the final call. Doesn't that protect against mistakes?
In theory, yes. But when you automate a judgment and present it with the appearance of scientific certainty—"the algorithm says adult"—it shapes how the human decision-maker thinks. They're more likely to defer to the machine, especially under pressure. The report warns against creating false certainty in inherently difficult decisions.
What makes age assessment so hard in the first place?
A 17-year-old who's been malnourished, traumatised, and exhausted from a dangerous journey may look older than their years. Their face doesn't match the reference images the AI was trained on. The algorithm has no way to account for what they've endured.
The Home Office talks about adults faking being children to access resources. Is that a real problem?
It happens, yes. But the data shows the bigger problem runs the other way—children are being classified as adults far more often than adults are falsely claiming to be children. The government's emphasis on "gaming the system" may be addressing a smaller issue while automating a larger one.
What would the organisations prefer instead?
They're not saying don't use AI at all. They're saying use it as one input among many, not as the decision-maker. Pair it with social worker assessments, legal representation, the ability to challenge the result. Build in safeguards. Don't let efficiency override accuracy when a child's safety is at stake.
What happens next?
The technology goes into further testing now, with a rollout planned for 2027. The question is whether the government listens to these warnings or treats them as obstacles to implementation.