Artificial intelligence has been embedded across the UK Home Office in near-total secrecy, reinforcing the so-called hostile environment for immigration — that is the central claim of a report by digital-rights group Open Rights Group (ORG), published overnight just after midnight UK time and reported by The Guardian. The Home Office disputes that characterisation, saying AI “does not replace or automate the role of human decision-makers” and that every asylum decision is made by trained caseworkers.
The report, titled Automated Discrimination: Unpacking the Digital Hostile Environment, draws on published Home Office material, freedom-of-information responses and parliamentary questions. According to ORG, AI now plays a role in detecting potential sham marriages, assessing the ages of asylum seekers, searching the phones of new arrivals, live facial recognition and AI-powered drones and satellites at borders, summarising asylum interviews, and triaging cases through a tool called Identify and Prioritise Immigration Cases (Ipic) — the workings of which, the group argues, are opaque.
What the report says, and what the department says back
ORG’s argument is less about any single tool than about the pattern: that these systems were trialled “in near-total secrecy”, that officials routinely invoke national security to stonewall FOI requests and redact disclosures, and that successive governments have treated complex human cases as a technical problem to be optimised. “Relying on AI to decide who gets to stay here and who gets thrown back into danger is a deeply alarming move,” the report states. It has cross-party backing from Labour’s Bell Ribeiro-Addy, the SNP’s Pete Wishart and the Greens’ Siân Berry, and the group is calling for a statutory ban on uses that risk human rights violations, a public register of government AI tools, and stronger freedom-of-information law.
ORG’s earlier work on the asylum caseload documents the concrete systems already in the pipeline: an Asylum Case Summarisation tool built on ChatGPT-4 that condenses interview transcripts, credited by the department with saving 23 minutes per case, and an Asylum Policy Search tool that summarises country guidance documents, credited with 37. ORG’s objection to those numbers is that speed is the wrong metric for decisions that can send someone back to persecution.
The Home Office, for its part, says AI is “just one of the tools at our disposal to protect our borders”, used only to help caseworkers navigate interview records and policy information, monitored through existing quality-assurance processes, with updated evaluations committed for publication. The department has not, per Ribeiro-Addy, published the documents that would show how the tools are evaluated or why “some of the most vulnerable people, with the least power, should bear the greatest risks.”
Both things can be true at once: the tools may well be assistive, and the scrutiny may still be inadequate. What is not in dispute is the asymmetry — a department deploying systems across marriage checks, age assessments, phone searches and border monitoring knows a great deal more about how they perform than the public or the applicants themselves.
Why a rights group’s report should be read as a report, not a verdict
ORG is an advocacy organisation, and “Automated Discrimination” is campaign literature as much as research — the Guardian’s coverage treats it as a claim, and so should readers. The countervailing case is real enough: an asylum backlog that stood above 62,000 initial decisions at end-September 2025, caseworkers making genuinely life-changing calls with limited time, and summarisation tools that plausibly free up human attention rather than replace it. The Hostile Office project by the Migrants’ Rights Network has documented the parallel digitisation of the border — eVisas, contactless crossings, biometric screening — for years, arguing migrants have become “a testing ground for the State’s digital surveillance plans”. Three organisations with a stake, one department with a transparency deficit, and no independent audit of any of it: that is the actual state of play.
The New Zealand angle is not hypothetical. The Government Communications Security Bureau’s surveillance-expansion push drew exactly the kind of scrutiny-first questions ORG is now asking of Whitehall, and our analysis of AI’s hidden costs in the NZ public sector makes the same core argument in a local key. Immigration systems are where governments reach for automation first — the stakes are per-case rather than systemic, the clientele lacks political leverage, and “process the backlog” is an irresistible slogan. The UK shows where that path leads when the audit function does not keep pace: not necessarily bad decisions, but an accountability gap wide enough that neither the public nor parliament can tell. The honest takeaway from the ORG report is not “AI decides asylum claims” — it is that nobody outside the Home Office currently knows whether it does, and the department has resisted the instruments (a public register, FOI reform, published evaluations) that would let anyone find out.
