Artificial intelligence (AI) has been integrated into a wide range of operations across the United Kingdom’s Home Office in relations to migration and asylum in “near total secrecy,” according to a new report published last week by the Open Rights Group (ORG), a UK-based digital rights organization. Watchdogs have warned that these systems are often inaccurate.
How Has AI Been Used Across Procedures?
The report details two avenues of integration: uses of AI in the immigration system and in border surveillance.
In the system, this includes summarizing interview transcripts using large language models (LLMs) through an algorithmic tool called ‘Identify and Prioritize Immigration Cases’ (IPIC), and operating the Electronic Monitoring Review Tool (EMRT), an automated system that provides decision recommendations by generating “harm scores” for those wearing ankle tags, before being allowed to transition to non-fitted devices.
Privacy International, a charity working to “to protect democracy” from governments and corporations’ use of technology for exploitation, filed a complaint over the uses of IPIC and EMRT in 2025, arguing that they “appear to [be] being used in immigration enforcement operations to assist or even replace human decision-making” adding that EMRT’s “harm scores” amount to “decisions made solely by automated processing.”
There is also the Atlas platform that digitizes casework and notifies applicants about specific milestones in their case’s process. Kaelynn Narita, Research and Policy Manager at the grassroots organization the3million, says this suffers from “patchy, dysfunctional and heavily automated features” that produced “76,000 misrecorded” entries “or data errors,” some of which have actually “ruined people’s lives.”
The report also touches on the new trial of AI face recognition to determine the age of asylum seekers, despite an “error margin of around 2.5 years.” There is also a possibility that technology to screen electronic devices found on an “illegal entrant” will use AI.
In border control, the report details aspects of the government’s plan to build an automated border for “seamless” and “frictionless” border crossings. This includes automated border crossing with an AI-based face recognition software, live face recognition to identify if arriving migrants have a deportation order, AI-powered drones and surveillance towers to “continuously scan the sea horizon,” and an “advanced risk assessment programme” called ‘Cerberus’ that would “collate passport and visa data, records from criminal databases and watchlists” to automatically provide actionable insights aimed at preemptively computing “risks” before the border.
Human Decisions Cannot Be Replaced
Though the UK Home Office has continuously stated that AI will not replace human judgment in asylum decisions, ORG’s report highlights the government’s reluctance to release information on where the technology comes from, its accuracy, and where exactly it is used, citing national security. The only publicly available information is that in some respects, OpenAI’s GPT-4 model is utilized.
An investigation published by British newspaper The Independent, highlighted that the AI tool used to speed up the processing of asylum cases by summarizing interview transcripts failed basic accuracy tests. Members of Parliament (MPs) and charities alike protested to the wider rollout of these systems. Nick Beales, at the charity Refuge and Migrant Justice, said ministers should “ensure that AI is never used as a substitute for human review of these life and death decisions.”
Home Office: Further Rollout to be Monitored
A Home Office spokesperson reiterated that “Asylum decisions are made by trained decision-makers, not artificial intelligence.”
“We regularly monitor the performance of these tools alongside our established quality-assurance processes.”
According to Infomigrants, the department has also committed to publishing updated evaluations of the tools.
Sara Alsherif, program manager at Open Rights Group, stressed that migration remains a “human phenomenon and experience, not a technical problem that can be fixed by artificial intelligence.”
