High-Risk AI in Migration, Asylum and Border Control

Domain 7 of Annex III AI Act covers AI systems used by migration, asylum and border control authorities. The Commission guidelines of 19 May 2026 connect this closely to existing European systems for border control and migration administration.

For the general framework, see the main article on the Article 6(3) filter. For all domains, see the hub overview.

The Four Use Cases of Domain 7

• Point 7(a) Polygraphs and similar tools in migration context
• Point 7(b) Risk assessment of persons seeking to enter or stay in the EU
• Point 7(c) Assistance in examining asylum, visa or residence applications and associated complaints
• Point 7(d) Detection, recognition or identification of persons in migration or border control context (excluding travel-document verification)

Interaction with Existing Systems

The guidelines explain that many AI applications in this context are integrated with European border systems:

• Schengen Information System (SIS) for alerts and refused entry
• EES (Entry-Exit System) for border control
• ETIAS for travel authorisation from visa-free countries
• EURODAC for asylum applications and biometric identification

AI integrated into these systems for decisions on persons falls under the relevant use cases of Point 7.

Use Cases in Detail

Point 7(a): Polygraph-Like Tools

High-risk:

• AI interview support at migration authorities or border guards assessing answer consistency
• Modern lie detection in asylum interviews

Point 7(b): Risk Assessment for Admission

High-risk:

• AI scoring ETIAS applications on risk
• AI ranking visa applications by suspicion
• AI predicting human smuggling or trafficking risk for individuals

Filter possible:

• AI generating statistical reports at aggregate level without individual assessment

Point 7(c): Assistance in Examination

High-risk:

• AI automatically linking country-of-origin information to asylum claims
• AI conducting language analysis to verify origin
• AI checking identity claims for internal consistency across complete asylum files

Filter possible:

• AI placing documents into fixed folders (identity documents, travel itinerary, supporting evidence)
• AI verifying documents for authenticity markers without candidate assessment

This example comes directly from the guidelines as an illustration of the narrow procedural task filter. Read the details in the main article on the filter.

Point 7(d): Identification in Migration Context

High-risk:

• Facial recognition at borders for 1-to-many identification (other than travel-document verification)
• Biometric identification in reception centres

Outside scope:

• Pure passport control where the traveller presents themselves (1-to-1 verification)

Sector-Specific Pitfalls

Pitfall 1: Schengen and AI Act Together

Many border systems are governed by EU regulations (SIS, EES, ETIAS, EURODAC). The AI Act sits on top. For Member State implementation this means border, immigration and customs authorities must be able to demonstrate both their sectoral legal basis and their AI Act compliance.

Pitfall 2: Vulnerable Groups

Asylum seekers are often in a vulnerable position with limited possibility to challenge automated decisions. Human oversight under Article 14 requires the human assessor to have sufficient room to deviate from AI output.

Pitfall 3: Fundamental Rights Take Centre Stage

For public authorities, Article 27 mandates a FRIA (Fundamental Rights Impact Assessment) for high-risk AI in this context. This touches non-discrimination, right to asylum and non-refoulement.

What to Do