Read more:
How does the Home Office use AI?
It’s difficult for researchers to determine exactly how the Home Office uses AI. Freedom of information requests have been stonewalled on national security grounds, the report’s author Fabio Chiusi explains.
“We don’t know enough about how the government actually uses AI in the immigration system and elsewhere,” he told Big Issue. “And we won’t be able to know enough unless we aggressively act on transparency laws and the very notion of ‘national security’, which is too vague and too often weaponised as a mere tool for opacity and secrecy, even when it should obviously be balanced with disclosures in the public interest.”
Keep reading the journalism that powers our mission
Register free to continue and learn more about the stories, campaigns and the support we provide, helping people facing poverty across the UK.
But despite these barriers, the report manages to document several uses of AI and automated decision-making in the immigration system.
One is summarising asylum interviews, condensing lengthy transcripts for caseworkers to analyse. Another is the Identify and Prioritise Immigration Cases (IPIC) tool, which assesses the removability of people subject to immigration enforcement and recommends actions including detention and deportation. (The Home Office disputes that IPIC is an AI tool, describing it instead as a rules-based system).
A third use is facial age estimation, which the Home Office plans to roll out in 2027 to help determine whether people seeking asylum are adults or children.
A Home Office spokesperson said that AI is “just one tool” to protect borders.
“Asylum decisions are made by trained caseworkers, with AI tools used only to assist in navigating interview records and policy information. They are not used in isolation and do not make decisions on claims,” they said.
“We regularly monitor the performance of these tools alongside our established quality-assurance processes, and have already committed to publishing updated evaluations on their roll out.”
But the risks of some of these tools are already apparent. An evaluation of the Home Office’s own AI trials, conducted between May and December 2024, found that nine percent of summaries produced were inaccurate or missing information. Almost a quarter of caseworkers involved in the trial (23%) said they were not fully confident in the summaries.
The Home Office insists that a human is always involved in the process. But ORG report author Fabio Chiusi cast doubt on the efficacy of this check.
“Multiple civil society and journalistic investigations showed that automation bias is alive and well, even when a human is effectively ‘in the loop’ of an automated decision,” he said.
“Humans tend to delegate respnsibilities to machines and trust their judgments. How can the government ensure that this won’t happen again and again with AI-powered tools, especially in the absence of strict and effective regulation?”
The Home Office has also acknowledged that its facial age estimation technology can vary in accuracy depending on a person’s ethnicity and skin tone.
Even the most accurate systems have an error margin of around 2.5 years when estimating whether someone is aged between 16 and 18 – precisely the age range in which the Home Office wants to use them.
The consequences could be severe, added Rosa Curling, co-executive director at Foxglove, a legal non-profit dedicated to fighting big tech.
“The Home Office obviously deals with some of the most vulnerable communities in our society,” she told Big Issue. “Sadly, we’ve seen through history that quite often government will test these technologies on the most vulnerable members of our society.”
She said companies developing the technology had themselves acknowledged differences in its accuracy across demographic groups.
“The Home Office has bought that equipment knowing that, and is suggesting it’s used on unaccompanied minors to determine whether they are over 16 or over 18. If you get that decision wrong, it has huge implications for that minor.”
If a person is determined to be an adult, they lose accommodation and legal rights that minors get.
The report also highlights problems with another automated system, designed to identify potential sham marriages.
Introduced in 2015, the tool adopted machine learning in 2019 and remains in use. It gives couples a red or green light, flagging some for further investigation based on six undisclosed risk factors.
An internal Home Office evaluation found that people from Albania, Greece, Romania and Bulgaria were disproportionately flagged. But the report released today says there is no evidence that people from these nationalities are more likely to be involved in sham marriages.
What about other government departments?
The government has big plans to roll out AI tools across public services. As prime minister, Keir Starmer declared an intention to “mainline AI into the UK’s veins”, hoping it could help do everything from fixing potholes to marking homework. His successor Andy Burnham has appointed the first minister for AI, and pledged to continue the rollout.
Chiusi calls this mindset “AI-solutionism”: a “complete ideological surrender of top government officials, including successive PMs from both Labour and Conservatives, to blind faith in AI and its powers to solve everything.”
“AI is already being widely adopted across government functions, way beyond the immigration system,” he warns. “If AI is not already there, it most likely will be.”
The Big Issue has previously reported on how automated decision-making in the Department for Work and Pensions is impacting claimants.
And just this week, policy documents from the Home Office revealed that the government is preparing to deploy facial recognition technology at transport hubs and public events, as part of a wider immigration crackdown.
For Chiusi, this points towards a future in which people are increasingly monitored and treated as potential threats.
“Even assuming they can be properly regulated and deployed, these technologies are intrinsically problematic, as they realise a securitised world in which everything and everyone is treated as a suspect of potential crimes,” he says.
“Democracies should not be based on a notion of security inspired by an AI-powered update to Bentham’s Panopticon, which – in the words of Home Secretary, Shabana Mahmood, means that “the eyes of the state can be on you at all times”.
Concerningly, this would represent an ideal outcome, rather than a dystopia, for the Home Secretary.”
The panopticon was an 18th-century prison design conceived by philosopher Jeremy Bentham, in which prisoners could never know whether they were being watched. The idea was that the possibility of constant surveillance would compel them to regulate their own behaviour.
As AI spreads through government, Chiusi warns that this kind of surveillance risks becoming normalised.
But he remains hopeful – as long as we don’t, as tech billionaires would have us do, see the roll-out of AI as inevitable.
“It is human minds and collective movements, rather than technologies alone, that actually shape the future,” he says. “If today we are surrounded by technologies for automated abuse, discrimination and control it is because somehow – with piles of cash, lobbying, and mountains of hateful lies – we left this crucial imaginative function to billionaires, tech ‘gurus”’and populist politicians.”
“Do you want a better future, one in which the environment is not hostile for anybody? Then start imagining it, and pretend it becomes real.”
Do you have a story to tell or opinions to share about this? Get in touch and tell us more.
Change a vendor’s life.
Buy from your local Big Issue vendor every week – and always take the magazine. It’s how vendors earn with dignity and move forward.
You can also support online:
Subscribe to the magazine or support our work with a monthly gift. Your support helps vendors earn, learn and thrive while strengthening our frontline services.
Thank you for standing with Big Issue vendors.