How Does the AI Act Apply to Migration and Border Control AI?

Which migration and border functions are listed?

Annex III point 7 lists four kinds of use, each of which must be tested on its own terms. They include polygraphs and similar tools to detect the emotional state of a person, systems to assess the risk, including a security risk, irregular immigration risk or health risk, posed by a person who intends to enter or has entered the Union, systems to assist competent authorities in examining applications for asylum, visas and residence permits and associated complaints as to eligibility, and systems to detect, recognise or identify natural persons in the context of migration, asylum and border control management, other than the verification of travel documents. A system that performs any of those functions is a candidate for high-risk classification, and a system that performs more than one must be assessed for each. The record should describe the function, the data, the output and the decision it influences, citing the Annex point for each function, rather than describing the system only as a border tool.

What must be screened under Article 5?

Start from the function, not the label border tool. Each function is tested on its own against Article 5: does it identify a person by biometric data, does it categorise a person by biometric data in a way that infers a sensitive attribute, does it infer emotion during an asylum or visa examination, does it score a person across contexts, or does it assess risk from profiling or characteristics in a way the prohibition covers? The inputs, the inference and the decision that follows must all be in the record, because a prohibition turns on what the system does, not on how the supplier describes it. Where a function sits close to a prohibition, the Article 5 text, the Article 3 definitions and the Commission guidelines on prohibited practices must be read together, with equality and non-discrimination law in view. A prohibited function must not move on to conformity work, and a borderline function needs qualified advice before deployment, with the advice and the facts it relied on kept with the file.

For the broader regulatory framework, see finance, insurance and critical infrastructure.

How does restricted registration work here?

Registration for high-risk systems in Annex III points 1, 6 and 7 in migration, asylum and border control management sits in a secure non-public section of the EU database under Article 49(4), including only specified information, with access limited to the Commission and the national authorities referred to in Article 74(8). A provider or public authority deployer must therefore identify the Annex point and area, enter only the specified items, and use the secure section. Deployer registration under Article 49(3) applies to public authority deployers, with Section C information likewise limited to the specified points in the restricted section. The restricted treatment reflects the sensitivity of the data and the vulnerability of the people affected, and it does not remove the duties to document, to assess fundamental rights impact where required, to keep logs and to exercise oversight. The underlying compliance work must be as complete as for a public registration, even though the entry itself is restricted.

What about large-scale IT systems?

AI systems that are components of the large-scale IT systems established by the legal acts listed in Annex X, in the area of freedom, security and justice, are subject to the transitional rule in Article 111(1): systems placed on the market or put into service before 2 August 2027 must be brought into compliance by 31 December 2030, without prejudice to the Article 5 prohibitions. Annex X lists the Schengen Information System, the Visa Information System, Eurodac, the Entry/Exit System, the European Travel Information and Authorisation System, the European Criminal Records Information System for third-country nationals, and the interoperability framework, each established by its own legal act. A system that is a component of one of those systems should be assessed under its own legal act together with the AI Act, with the transition, the evaluation requirements and the data protection position checked together. A system that is not a component of an Annex X system should not borrow the 2030 transition: its application date follows its own Annex III route, and its compliance plan should be built to that date.

What oversight and safeguards are needed?

Oversight here has to work under pressure, with people whose circumstances vary widely. The reviewer needs time, the full case information, and real authority to depart from the system output, including where the output points one way and the person's evidence points another. Information for the person affected must be given in terms they can understand and in a language they can access, covering that an AI system was used, what it did, and how the decision can be challenged under the procedural law that governs the application or examination. Safeguarding routes matter as much as the technical ones: interpreters, legal assistance and protection referrals should be planned alongside the system, with particular care for children, victims of trafficking and people seeking protection. Outcomes should be tracked for consistency across groups, for how often outputs are followed or overridden, and for what challenges reveal, with each review, decision and outcome recorded.

Continue with law enforcement AI systems.

What does this guide not decide?

Nothing in this guide settles a real migration or border case. Whether a given system is prohibited, high-risk or authorised, and whether any application, examination or border decision was lawful, depends on the function, the Article 5 exceptions, the Annex III listing, the Annex X legal acts where relevant, and the national and Union migration and data protection law applied to the specific facts, with qualified advice. The guide is written for the EU position only and describes no other jurisdiction's law. Before a migration or border system is used, the prohibition, classification, registration, supervision and procedural questions should be confirmed with the competent authorities and legal advisers, and the assessment should be kept with the deployment record.

Frequently asked questions

Continue through this AI Act cluster

• finance, insurance and critical infrastructure
• law enforcement AI systems
• justice and democratic processes AI
• how AI Act transitions work for systems already on the market
• what is public and what is restricted in the EU database
• high-risk AI systems