The Home Office has embedded artificial intelligence across its immigration apparatus — from detecting potential sham marriages to determining asylum seekers' ages, searching phones of new arrivals, summarising asylum interviews, and prioritising cases through its opaque Identify and Prioritise Immigration Cases (Ipic) tool. The scope is sweeping. The disclosure is not. A new report by Open Rights Group, titled Automated Discrimination: Unpacking the Digital Hostile Environment, documents each deployment using FOI requests, MP questions, and the limited crumbs the department has published voluntarily. The transparency problem is structural, not incidental. Officials routinely invoke national security to stonewall FOI requests and redact disclosures, according to the report. This is not a government deploying experimental technology and inviting public scrutiny. It is a government deploying experimental technology on the people least able to challenge it and then classifying the details. Bell Ribeiro-Addy (Labour), Pete Wishart (SNP), and Siân Berry (Green) have all backed the report's cross-party call for accountability. The surveillance layer is its own concern. Live facial recognition, AI-powered drones, and satellites are being used for border surveillance. These are not speculative future applications — they are operational. The report connects these capabilities directly to the hostile-environment doctrine first articulated by Theresa May as home secretary in 2012, arguing that algorithmic systems do not neutralise the policy's intent but encode and scale it. Open Rights Group's central argument is that AI in immigration is not neutral. When the political driver behind deployment is anti-immigration rhetoric, the technology inherits and amplifies that bias. The report warns against reducing complex human cases to technical problems. It calls for regulation of government AI, a ban on undemocratic uses (including ChatGPT tools in asylum processing), stronger FOI laws, a narrower definition of national security, and a public register of all government AI tools. The Home Office response follows a now-familiar template: AI is a tool, not a decision-maker; trained caseworkers make asylum decisions; AI only assists in navigating interview records and policy information; quality assurance processes exist; updated evaluations will be published. The statement does not address the FOI stonewalling, the national security invocations, or the absence of a public AI register. It does not engage with the surveillance deployments. Ribeiro-Addy put the accountability gap plainly: the department has not explained how it will ensure accountability when these tools fail, nor justified why the most vulnerable people with the least power should bear the greatest risks. Sara Alsherif of ORG framed it as a category error: migration is a human phenomenon, not a technical problem. The pattern here is not unique to the UK Home Office, but the combination — secretive deployment, national-security classification of mundane operational details, deployment on populations with no political leverage, and a rhetorical framework that treats AI as merely assistive — represents a governance failure that compounds with every new tool added. The question is not whether AI will be used in immigration; it is whether any democratic mechanism exists to constrain how.
