Home Office knew AI age checking technology for migrant children was flawed — but will roll it out anyway

The Home Office knew an AI tool to check the age of small boat migrants was flawed but pressed for its rollout anyway. Independent can reveal.

According to a leaked report that the Ministry of Internal Affairs tried to hide, artificial intelligence-supported technology that estimates a person’s age based on facial features mistakenly evaluates young people as adults.

The confidential report also found the technology was least accurate when trying to age migrants from countries such as Eritrea and Sudan, where the largest number of small boat migrants come to the UK, amid accusations that the technology was “based on racial bias”.

He warned that error rates were particularly high for female child migrants from sub-Saharan Africa, with an average of 4.6 years, meaning a 14-year-old girl could be predicted to be an adult.

The report, prepared in April 2025 by officers responsible for the biometrics programme, also warned that the technology could be less accurate for people experiencing visible aging due to “travel stress”.

Separately, an audit by Lighthouse Reports of publicly available data on the government’s chosen AI provider showed that its technology misclassified more than a third of 16-year-olds as adults and misjudged 70 percent of cases in some tests.

The Home Office’s scientific advisors have spoken out for the first time, telling this publication that they felt the government was rushing to adopt AI technology for political reasons and chose not to consult them to avoid criticism.

Professor Tim Cole, from UCL’s Institute of Child Health, suggested the department was pursuing the technology despite being aware it was “horribly flawed”.

Ministers announced last month that the Centenary Age Estimation (FAE) will be used by immigration officers at the border from next year to “weed out false claims by those arriving in small boats disguised as children”.

But despite knowing the serious problems, the Home Office said its “cost-effective” age assessment method showed “promising performance and accuracy”.

An investigation by Independentin collaboration Via Lighthouse Reports and WIRED found that the government’s own evaluation of the FAE found that the technology predicted that 17-year-olds would be over 18 on average.

The technology said that in one in 20 cases, a 17.5-year-old falls outside the 14 to 22.5 age range. It also performed worse in teenage girls.

Meanwhile, 60 organizations sent an open statement today. letter It called on the Home Office to halt the rollout of FAE, warning that the technology had “failures and discrimination” and was “not accurate at the critical 16-18 age limit”.

Martha Dark, co-executive director of Foxglove, the nonprofit organization that coordinated the letter, said children should not be test subjects for “experimental technology that introduces inaccuracy and racial bias,” adding: “Errors from these tools can have serious consequences: vulnerable children can be forced to go to adult detention centers, alone.”

The Facial Age Estimation Performance Test report, which the Home Office declined to disclose under a freedom of information request but was later leaked to Lighthouse Reports, evaluated seven commercially available algorithms, including those ranked top by the National Institute of Standards and Technology (NIST) in the US.

Although the Ministry of Internal Affairs report did not specify which artificial intelligence algorithms were examined, it was revealed that the technology used in the rollout would be provided by the German company Cognitec Systems.

The Home Office said the conclusions reached in the FAE test report did not relate to the active procurement of services from any provider.

Separately, the Home Office said in response to a freedom of information request that Cognitec was among seven suppliers of artificial intelligence technology used for internal testing.

Cognitec’s algorithm was reviewed by NIST and the findings were published in May. Analysis of publicly available data shows that the Cognitec tool incorrectly classified more than a third of 16-year-olds as adults when tested on US visa application photos.

Data analysis found that this rate rose to almost 70 percent when the AI ​​system was tested on lower quality border crossing photos.

A spokesperson for Cognitec said the algorithm’s bias is low compared to other algorithms and that they are working to reduce it. They added that issues with demographic differences affect all vendors and are mostly related to image quality issues.

The Home Office has not published details of exactly how the AI ​​technology will be used by border authorities, but the internal report discusses using an age threshold that would give the system a margin of error.

For example, if the age threshold is set at 20, anyone estimated to be 20 years old or younger is classified as a minor.

Increasing the age threshold will reduce the risk of children being treated as adults, but will also increase the number of adults treated as children. The report warns that if the threshold is set at 25, this will inevitably lead to a “large” number of adults being classified as minors.

A contract for FAE technology has been awarded £322,000 to IT provider Akhter Computers, with the algorithm provided by Cognitec.

The move to use FAE on asylum seekers is unprecedented. While FAE is increasingly being used globally across online platforms and retail environments to verify that children cannot access adult-only content or purchase age-restricted products such as alcohol and cigarettes, there are no known cases of it being used in an immigration context.

The lack of accuracy of information on sub-Saharan citizens poses a major problem for the Ministry of Home Affairs; because asylum seekers from Eritrea and Sudan have become the top nationals making small boat journeys across the Canal, surpassing migrants from Afghanistan and Iran as the top crossers in 2024.

More than a third (36 percent) of people age-assessed in 2025 were from Sub-Saharan African countries, according to government data. The number of asylum seekers from sub-Saharan Africa subject to age assessments has increased steadily over the years; It increased from 163 people in 2015 to 2,331 people in 2025.

Tests of the AI ​​algorithms were carried out on facial images taken from the Home Office’s own immigration databases, specifically records of visa and residence permit applicants, but not asylum seekers.

The Home Office said internal tests were carried out in line with legal advice and the images were not shared externally.

Before it was disbanded in July 2025, the Home Office had a panel of scientific advisors to guide age assessment methods, called the Age Estimation Science Advisory Committee (AESAC).

Professor Cole, a former member of the committee who was awarded a CBE in 2025 for services to medical statistics, said he felt the committee had been held back by the Home Office because it did not want members to examine the move to artificial intelligence.

He said: “We wanted to highlight the inadequacies of centennial age estimation but we were not given the opportunity and the committee was subsequently closed down, so I suspect we will be quite damning about it.

“Their claim was that our expertise was very specialized and not relevant to facial age estimation. In reality, what facial age estimation does is try to interpret age-related changes in appearance, which are fundamentally biological, and our background is biological growth and aging.”

He said he had seen the Home Office’s internal review of the FAE and offered to carry out a statistical review but was rejected.

Denise Syndercombe-Court, professor of forensic genetics at King’s College London, said she believed the Home Office disbanded the committee because they feared it would slow the spread of FAE: “This was probably because they believed they had a cost-effective technological alternative in the form of FAE. It needed to be rapid to meet perceived urgent political demands and we were not asked for our opinion.”

He added: “I think this should be examined scientifically and statistically before the technology is used. I have no problem with that as long as it’s fair and proportionate to people’s exposure to it – and I’ve seen no evidence that that’s the case.”

The Home Office said they commissioned the National Physical Laboratory to carry out an independent review of the tests and trial reports produced when implementing FAE. They added that the FAE needed different areas of expertise to provide advice.

A spokesman said: “Robust age assessments are a vital tool in securing borders and protecting children. “We have stringent processes for verifying an individual’s age and are working to streamline these by testing fast and effective Facial Age Estimation technology.

“This groundbreaking companion tool is designed as an additional source of information for immigration officers and does not replace or override human judgment.”