Home Office Awareness of AI Age Checks’ Flaws Doesn’t Deter Rollout
Home Office knew AI age checks – Exclusive findings reveal that government advisors labeled the AI age assessment tool as ‘severely flawed’ in internal evaluations, yet the Home Office proceeded with its deployment. The technology, designed to estimate the age of small boat migrants, has been criticized for misjudging teenagers as adults due to inherent biases in its algorithms. These concerns were highlighted in a leaked report that the Home Office attempted to suppress, which underscores the system’s potential to misidentify children as adults, especially those from certain regions.
Algorithmic Bias and Error Rates Highlight Concerns
The AI system, which analyzes facial features to determine age, was found to be most unreliable when assessing migrants from Eritrea and Sudan—countries with the highest numbers of small boat arrivals to the UK. This inaccuracy has raised alarms about the tool’s ‘baked-in racial bias,’ as it disproportionately affects children from Sub-Saharan African nations. For instance, the report indicated that female child migrants from these regions were often misclassified by up to 4.6 years on average, leading to situations where a 14-year-old girl might be deemed an adult.
“Errors by these tools could have serious consequences: vulnerable children being forced, alone, into adult detention centres.”
Additionally, the technology struggles with individuals exhibiting visible signs of aging due to the physical and emotional toll of travel. This issue further compounds the risk of incorrect age classifications, potentially impacting decisions about asylum claims and legal status. The report, authored by civil servants in April 2025, evaluated seven commercially available algorithms, including those rated highly by the US National Institute of Standards and Technology (NIST). Despite this, the Home Office opted for Cognitec Systems’ algorithm, which has since come under scrutiny.
Political Motivations and Expert Criticisms
Scientific experts have criticized the Home Office for its haste in adopting the AI technology, suggesting that political pressures outweighed technical caution. According to the leaked findings, advisers felt the government prioritized rapid implementation to address public sentiment about migrant children making false claims. Professor Tim Cole of University College London’s Institute of Child Health emphasized that the system’s inaccuracies were well-documented, stating, “The department is pursuing this technology despite its glaring inaccuracy.”
The Home Office defended its decision, asserting that the AI method was deemed ‘cost-effective’ and ‘promising in performance and accuracy’ based on its internal assessments. However, the evaluation revealed that the system overestimated the age of 17-year-olds, often placing them in the adult category. In one in 20 cases, the tool incorrectly classified a 17.5-year-old as outside the 14 to 22.5-year range, with teenage girls facing even greater challenges in accurate age prediction.
Collaborative Investigation and Public Response
Following a joint investigation by *The Independent*, Lighthouse Reports, and WIRED, the flaws in the AI system have become more evident. The report, which the Home Office initially withheld from public access, was later leaked and scrutinized by the media. It revealed that the technology’s performance was subpar for young individuals, particularly those in the 16 to 18-year-old range, a critical demographic for immigration decisions.
In response to mounting pressure, 60 organizations have issued an open letter to the Home Office, urging a pause in the rollout of Facial Age Estimation (FAE). The letter warns that the system’s built-in inaccuracies and discriminatory tendencies could lead to serious misjudgments. Martha Dark, co-executive director at Foxglove, a non-profit organization involved in the letter, stated, “Children should not be test subjects for experimental tech that has baked-in inaccuracy and racist bias.”
Provider Selection and Public Data Analysis
While the Home Office report did not explicitly name the AI algorithms tested, subsequent revelations confirm that Cognitec Systems’ technology will be used in the rollout. The company’s algorithm was assessed by the NIST in May, with results showing that it misclassified over a third of 16-year-olds as adults. In some tests, the error rate reached as high as 70%, raising doubts about its reliability for critical immigration assessments.
Despite these findings, the Home Office maintained that the report’s conclusions did not reflect its active procurement process. The agency stated it had used Cognitec’s technology for internal testing but had not finalized a contract. However, the leaked data and subsequent analyses have exposed the system’s shortcomings, particularly its tendency to mislabel children as adults. This has sparked a debate over whether the Home Office is compromising fairness and precision for political expediency.
Ministers announced in June 2025 that FAE would be introduced at border checkpoints in 2027 to combat claims of children arriving as adults. Yet, the evidence suggests the system’s flaws were already known, and the decision to implement it was made without full consultation. Critics argue that the technology’s flaws could result in vulnerable children being detained alongside adults, undermining the protections meant for minors.
Call for Transparency and Reform
As the rollout of the AI system moves forward, calls for transparency and reform have intensified. The leaked report and public data analysis highlight the need for more rigorous testing and oversight before deploying such technology on a large scale. With the potential for significant errors, experts warn that the system could perpetuate injustices, especially for young migrants from regions already underrepresented in the system’s accuracy.
Meanwhile, the Home Office faces mounting scrutiny over its decision-making process. While it insists on the AI’s cost-effectiveness, the data paints a different picture. The combined efforts of *The Independent*, Lighthouse Reports, and WIRED have brought these issues to light, prompting questions about the agency’s commitment to fairness in its immigration policies. As the technology prepares for real-world application, its accuracy and equity will remain central to the debate over its use in assessing the age of asylum seekers.
