Government urged to abandon plans to use AI age checks for child asylum seekers

Concerns about accuracy of and transparency

More than 60 human rights and civil liberties organisations have called on the government to abandon plans to use AI to estimate the ages of asylum seekers, warning that the technology could wrongly classify vulnerable children as adults.

In an open letter to immigration minister Alex Norris, a coalition of 62 groups argued that facial age estimation systems are too unreliable and potentially discriminatory to be used in asylum decisions.

The organisations, which include Amnesty International, Human Rights Watch, Liberty, Electronic Frontier Foundation, Foxglove and Open Rights Group, have urged the Home Office to halt the planned rollout of the technology, which is expected to begin from 2027.

The system is intended to help immigration officials assess whether people claiming to be under 18 are likely to be children or adults.

Ministers have stressed that AI-generated estimates would be used alongside, rather than instead of, human judgement. Under current Home Office guidance, disputed cases are assessed using a "benefit of the doubt" approach.

Individuals are only treated as adults without further age assessment if two immigration officers independently conclude that their appearance and behaviour strongly suggest they are significantly older than 18.

Where uncertainty remains, individuals are treated as children and referred to local authorities for further assessment. However, campaigners say the safeguards do not address fundamental concerns about the technology itself.

Varying levels of accuracy

In their letter, the groups argue that facial age estimation systems have demonstrated varying levels of accuracy depending on ethnicity and skin tone. This raises questions about their suitability for assessing asylum seekers, many of whom come from ethnic minority backgrounds.

The organisations also pointed to what they see as a major limitation of the technology: its inability to reliably assess age among older teenagers, the very group for whom the system is intended to be used.

According to the coalition, government documents acknowledge that age estimation systems are least precise when assessing people aged between 16 and 18.

Margin of error

Campaigners say that margin of error could have significant consequences for young people seeking protection.

They also argue that asylum-seeking children may be particularly difficult for AI systems to assess because many have experienced trauma, conflict, malnutrition, sleep deprivation and hazardous journeys before arriving in the UK.

"The faces of children seeking asylum tend to prematurely age from trauma and violence," the groups wrote, adding that such experiences may make them appear older than their actual age.

The coalition questioned whether the technology has been adequately tested on individuals with those backgrounds and said there is insufficient evidence to show that it would improve decision-making.

Concerns over transparency

There are also concerns about transparency. The groups said the Home Office has published only limited information about how the systems were developed and tested.

They are seeking further details about the datasets used to train the technology, the methods used to evaluate its accuracy, and the safeguards available to challenge decisions.

The organisations have given the Home Office 21 days to answer a series of questions covering testing procedures, training data, appeal rights and the role AI-generated estimates would play in age assessments.

"We ask now that the Home Office take on this uphill task in good faith and release the information that is required to scrutinise the legality and ethics of this deployment of new technology on some of the most vulnerable people – and children – in our world," the groups said.