Summary

• Open Rights Group examined case-management, identity and surveillance tools across the immigration system.
• The group's conclusions draw on documents, information requests and interviews rather than evidence of fully autonomous asylum decisions.
• The Home Office says AI supports trained officials and is subject to monitoring and quality assurance.

A new investigation by Open Rights Group raises concerns about the transparency and governance of AI-assisted tools used or trialled in UK immigration and border administration, including systems for case prioritisation, identity assessment and surveillance.

Published on 27 October, the report draws on freedom of information disclosures, official documents, interviews and previous research. Its conclusions about discrimination, insufficient scrutiny and the risks of a digital hostile environment are the authors’ assessments, not findings formally admitted by government.

The Home Office says trained officials remain responsible for asylum decisions and that AI is used to assist rather than replace them. It also points to monitoring and quality assurance, although the researchers argue that public information remains insufficient to evaluate those safeguards independently.

Different systems, different consequences

The report examines a collection of technologies rather than a single automated decision platform. The functions under review include prioritising cases, summarising asylum interviews, analysing mobile phones, assessing age and helping identify suspected immigration breaches.

Case prioritisation can influence how quickly an application is considered without determining the final legal outcome. Its fairness depends on the criteria used, data quality and whether officials can identify and correct unexpected classifications.

AI-generated interview summaries pose another problem: fluent text can omit qualifications, introduce inaccuracies or give undue prominence to particular statements. Human review may identify such mistakes but cannot be assumed to catch every defect.

An earlier Open Rights Group investigation focused specifically on generative AI in asylum processing. The new report is broader, encompassing border technologies and questions about procurement, public information and oversight.

The Home Office has also published information about planned facial recognition activity. Techopia examined those plans on 26 October; the newer ORG analysis is distinct in both scope and the evidence it gathers.

Border surveillance and evidence gaps

Open Rights Group examines facial recognition, drones, satellite systems and proposals involving predictive surveillance. Their capabilities and deployment status vary, so it would be inaccurate to describe every tool mentioned as fully operational.

Surveillance technology may provide information to officials, but the accuracy and significance of its outputs depend on environmental conditions, detection criteria and how information is matched against other records.

Wrongly associating a person with a record or misclassifying an image may have significant consequences if it prompts further enforcement action. Meaningful assurance requires information about error rates, human review and the process for challenging an incorrect result.

The organisation says requests for information have frequently encountered incomplete responses or redactions, in some cases on security grounds. Some operational secrecy can be justified, but public authorities must still consider what they can disclose about purpose, legal basis and accountability.

Documentary evidence may confirm a programme’s existence without proving its effectiveness. Neither the fact that a system has been trialled nor a supplier’s description of its intended functionality establishes reliable performance in practice.

Government oversight and contractor dependence

The investigation raises questions about the use of overseas technology suppliers and the ability of public bodies to inspect proprietary systems. Supplier nationality alone does not establish a security or legal breach, but contractual access and transparency can influence oversight.

Public authorities need to establish who operates each technology, how updates are controlled, what information may be transferred and which party is responsible when a problem arises. Contracts should make investigation and audit possible even when software is not developed in-house.

Software systems may also create switching costs as staff practices, historical data and other applications become dependent on them. That can make independent assessment and data portability important long before a contract reaches its renewal date.

Open Rights Group recommends stronger transparency records, information-access rules and controls on high-risk uses. Those are advocacy recommendations and should not be presented as existing legal obligations or as an agreed government reform programme.

The Home Office’s position that trained officials make asylum decisions is an important limit on what the report establishes. It does not eliminate the possibility that automated suggestions influence judgement, but the available evidence does not show AI autonomously granting or refusing asylum.

The published report identifies systems and disclosure gaps that its authors believe merit scrutiny. Independent evaluation will require fuller technical evidence about what is deployed, how performance is tested and how individuals can challenge errors.

Different uses also raise different legal and ethical questions. Summarising a document is not equivalent to matching a face against a watchlist, and an appropriate oversight regime must reflect the sensitivity and consequences of each task.

Fairness testing requires more than an overall accuracy figure. If errors are unevenly distributed across groups or operational settings, a superficially strong average can mask serious impacts on particular individuals.

Where authorities rely on commercial tools, they should be able to explain whether staff can inspect source material underlying an automated recommendation. Otherwise, an affected person may struggle to identify why a particular administrative action occurred.