- The UK government intends to use facial age estimation (FAE) AI to determine the ages of asylum seekers at the border, a first-of-its-kind application.
- An internal government report found that the FAE systems tested are prone to misidentifying children as adults by several years.
- The AI technology exhibits significant bias, performing worse on individuals with darker skin tones and specific ethnic backgrounds, including those from Africa and the Middle East.
- Misidentifying a child as an adult can lead to the loss of legal protections, placement in adult detention facilities, and increased risks of abuse and deportation.
- Despite the report’s conclusion that the systems are unreliable for high-stakes decisions, the government plans to proceed with their deployment.
- Human rights groups and legal experts criticize the plan, warning of severe human costs and potential legal challenges due to the technology’s known flaws and discriminatory nature.
The Plan: Using AI for Asylum Seeker Age Checks
Imagine arriving in a new country as a child fleeing war, only to have a computer misjudge your age and send you to adult detention. This could soon be the reality for asylum seekers in the UK.
The British government plans to use facial age estimation (FAE), an AI technology that scans faces to predict age, to help determine the age of asylum seekers at the border. While common online for age verification, this would be its first use for deciding the age of vulnerable individuals seeking safety.
Many asylum seekers lack identification documents due to fleeing their homes suddenly or coming from regions where official papers are scarce. Without documents, border officials must guess their age. Misidentifying a child as an adult can lead to severe consequences, including the loss of special legal protections and placement in harsher adult detention centers with higher risks of abuse.
The government claims the system will speed up processing and prevent adults from falsely claiming to be children. However, an internal report reveals the technology is flawed. It shows that the FAE systems tested frequently misidentified children as adults and exhibited bias against certain migrant groups.
Despite this internal evidence, the government intends to proceed, a decision criticized by human rights advocates as a dangerous gamble with people’s lives.
What the UK Government Knows: An Internal Report on Flawed AI
An internal, non-public report, leaked to journalists, details tests of several FAE technologies considered for border use. The findings are critical of the technology’s reliability.
The report indicates that FAE systems often misjudge ages by several years, with some children as young as 10 or 11 incorrectly classified as adults. While company names were withheld, the report notes that some providers have faced criticism for biased algorithms that perform poorly on individuals with darker skin tones. Others have sold faulty systems to law enforcement and educational institutions.
Crucially, the report highlights that the errors are not random but follow patterns. The systems are more likely to misidentify girls as older and show bias against people from specific ethnic backgrounds, making the technology discriminatory.
Government data shows that in 2025, the largest groups undergoing age assessments were from Africa and the Middle East. These are precisely the demographics the FAE systems performed worst on, meaning those most likely to be affected are also most likely to be misjudged.
The report concluded that the FAE systems were not reliable enough for high-stakes decisions like border age assessments. Nevertheless, the government has decided to proceed, drawing criticism from human rights and legal experts.
The Human Cost: Children Mistaken for Adults
An incorrect age assessment can drastically alter a person’s future. When a child is wrongly classified as an adult, they lose access to vital child protection services. They may be housed with unknown adults in detention centers, face prolonged detention, and risk deportation to unsafe situations.
Human rights organizations have documented cases of children placed in adult detention, reporting feelings of fear, isolation, and vulnerability. Some have experienced assault or developed mental health issues as a result.
The FAE system’s errors can be difficult for a misidentified child to challenge, especially if they do not speak English or have legal representation. Critics argue that even with human oversight, officials might overly rely on the AI’s assessment, failing to question its accuracy.
This technology also removes the nuanced human judgment involved in current age assessments, which are conducted by trained social workers and doctors. FAE replaces this careful evaluation with a quick scan, prioritizing speed over thoroughness.
As one advocate stated, “You cannot replace a conversation with a camera. A child’s life is not a math problem.”
Algorithmic Bias: Who is Most Likely to Be Misidentified?
The internal report clearly indicates significant bias issues within the FAE systems. This bias stems from how AI models are trained.
AI bias occurs when training data is not representative of all populations. If an AI is trained primarily on images of white Europeans, it will be more accurate for that group and less accurate for others. This is precisely what the report found.
The tested systems were more accurate for lighter-skinned individuals than darker-skinned ones, and more accurate for men than women. They were particularly inaccurate when estimating the ages of children from certain ethnic groups.
This is critical because asylum seekers are not a random global sample; they often come from regions affected by conflict or climate change, such as Africa, the Middle East, and South Asia. These are the very groups on whom the FAE systems perform poorly.
While companies were not named, similar systems have shown significant disparities. Research indicates commercial facial recognition systems can have error rates up to 35 percent for darker-skinned women, compared to less than 1 percent for lighter-skinned men. Age estimation systems exhibit similar problems.
This bias is not a minor technical flaw but a systemic issue that disproportionately harms vulnerable populations, reinforcing existing inequalities and hindering fair treatment.
From Online to Borders: The Expansion of Age Verification
Age verification technology is now common online, used by social media platforms and adult content sites. Many jurisdictions legally require it for certain websites.
These online systems typically function by analyzing a user’s selfie to estimate their age. Access is granted or denied based on the AI’s prediction.
However, the consequences of online age verification are vastly different from those at a border. While online, the worst outcome is restricted access to content. At a border, the stakes involve a person’s freedom and safety.
The UK government’s plan represents a significant shift from virtual to physical applications of FAE, marking the first time it will be used for a binding age determination at a border. This sets a potentially dangerous precedent, which other countries might follow, leading to wider use in airports, police stations, and schools.
Civil liberties groups warn this is a slippery slope, arguing that deploying a known-flawed system on vulnerable people for border control is unacceptable, even if it functions adequately for less critical online uses.
Global Use of Age Checks: Are They Ready for Border Decisions?
While governments worldwide use AI and facial recognition at borders, the UK’s plan to use FAE for asylum seeker age assessment is novel.
The US uses facial recognition at airports but not for determining migrant ages. Australia has considered age verification for social media but not border control. The UK’s initiative appears to be pioneering this specific application.
This global context highlights that FAE technology is still developing and not yet accurate enough for high-stakes decisions. Studies, like one from the National Institute of Standards and Technology in 2024, show that even the best age estimation algorithms have error rates of several years, with larger errors for children.
The UK government’s own report corroborates these findings, yet the plan proceeds. This suggests a prioritization of speed and efficiency over accuracy and fairness, with a willingness to accept risks to individuals’ lives.
Legal experts anticipate potential court challenges from asylum seekers misidentified as adults, arguing wrongful detention. Human rights groups may also contest the system’s compliance with the right to a fair age assessment.
Next Steps: Demands for Transparency and Accountability
The UK government has not disclosed which FAE system will be used, how it will be implemented, or why it is proceeding despite internal warnings. This lack of transparency is a major concern.
Advocates are urging the government to halt the plan for a comprehensive review. They call for independent testing of FAE systems, public release of results, robust safeguards against harm, and a clear process for asylum seekers to challenge AI-driven age assessments.
Some propose banning the technology entirely for this purpose, arguing that AI should not make life-altering decisions for vulnerable individuals, and that human judgment remains superior.
The government has yet to respond to these calls, but pressure is mounting. The investigative reporting has brought the issue to public attention, prompting questions from lawmakers and preparations for legal action by rights groups.
Currently, the plan remains on track. Starting next year, asylum seekers arriving in the UK may face AI facial scans, with the risk of children being wrongly placed in adult detention. The government is aware of these risks but is proceeding regardless.
The critical question is whether the government will heed its own experts’ findings. So far, the answer appears to be no.
Frequently Asked Questions
What is facial age estimation (FAE) and how will it be used for asylum seekers in the UK?
Facial age estimation (FAE) is an AI technology that analyzes facial features to predict a person's age. The UK government plans to use this technology at its borders to help determine the age of asylum seekers, particularly those without identification documents.
Why is using FAE on asylum seekers controversial?
The technology is controversial because an internal UK government report revealed it is flawed, frequently misidentifying children as adults and showing bias against certain ethnic groups. Critics argue this poses a significant risk to vulnerable individuals.
What are the risks if a child asylum seeker is misidentified as an adult?
If a child is misidentified as an adult, they can lose crucial legal protections, be placed in adult detention centers with harsher conditions and higher risks of abuse, face prolonged detention, and potentially be deported to unsafe situations.
Does the FAE technology have biases?
Yes, the internal report indicates significant bias. The systems are less accurate for individuals with darker skin tones and perform worse on people from certain ethnic backgrounds, including those from Africa and the Middle East, who are often a significant portion of asylum seekers.
What did the internal government report conclude about FAE?
The internal report concluded that the FAE systems tested were not reliable enough for high-stakes decisions, such as age assessments at the border, due to their inaccuracies and biases.
What are critics saying about the UK government's decision to proceed?
Critics, including human rights groups and legal experts, are calling the government's decision a dangerous gamble with people's lives. They argue that deploying a known-flawed system on vulnerable populations is unacceptable and could lead to severe human rights violations.
What are advocates calling for?
Advocates are calling for the UK government to pause the plan, conduct a full review, and ensure independent testing of the FAE systems. They also demand transparency, public release of results, clear safeguards, and a mechanism for asylum seekers to challenge AI-based age assessments.
