TL;DR

A report from Open Rights Group (ORG), covered by the Guardian on 27 October, says the Home Office has built AI into a wide range of immigration work with almost no public disclosure. ORG wants uses it sees as undemocratic banned, freedom of information rules strengthened, and a public register of every AI tool government uses.

Where the report finds AI

The report, titled “Automated Discrimination: Unpacking the Digital Hostile Environment”, lists AI in tasks such as flagging suspected sham marriages, judging how old asylum seekers are, and searching the phones of people who have just arrived. It also finds the technology at the border, through live facial recognition, drones and satellite imagery, and in casework, where it summarises asylum interviews and the country policy material used as evidence.

Cases are also sorted and assessed by a tool ORG describes as opaque, known as Ipic, short for “Identify and Prioritise Immigration Cases”. Its evidence comes from what the Home Office has published, from FOI requests and from MPs’ questions. ORG says officials routinely cite national security to refuse requests or redact what they release.

Three MPs from different parties back the report: Bell Ribeiro-Addy of Labour, Pete Wishart of the SNP and Siân Berry of the Greens. Ribeiro-Addy said the department has yet to set out who answers for it when the tools go wrong.

The Home Office response

The Home Office, through a spokesperson, said AI helps protect the borders but does not take the place of human decision-makers. In asylum work, trained caseworkers make every decision and AI only helps them move through policy material and interview records. The department says it monitors the tools and has promised updated evaluations of the rollout.

Context

The report links its concerns to the hostile environment policy Theresa May set out in 2012, when she was home secretary. ORG argues that AI built for an anti-immigration agenda is not neutral and can reproduce existing bias.

The asylum summarisation tool has drawn criticism before, as our coverage of expert checks on its summaries set out. This report widens the frame from one tool to the whole department.

Looking Forward

ORG’s sharpest demand is specific: a ban on using ChatGPT tools for asylum claims. In our view, the call for a public register is the one most likely to gain ground, because it asks for transparency rather than prohibition. For suppliers of public sector AI, it signals that procurement may soon come with disclosure duties attached, and that tools touching immigration decisions will draw the closest scrutiny.