Asylum, Migration and Border-Control AI Under the EU AI Act: Annex III Point 7

If you build or deploy AI that sits anywhere in the asylum, migration, visa or border-control pipeline, the default classification is high-risk. Annex III point 7 of Regulation (EU) 2024/1689 names this use area directly, and the obligations are heavier here than almost anywhere else — because the people affected are, by definition, in a vulnerable position and often cannot easily challenge a decision. This page sets out exactly which uses are high-risk, where the prohibited line sits, and what a public authority and a vendor each have to do.

The high-risk regime for stand-alone Annex III systems applies from 2 December 2027 under the Digital Omnibus, now adopted by Parliament and Council — one note to keep in mind, set out below: entry into force is pending Official Journal publication.

---

What Annex III point 7 actually covers

Annex III point 7 lists four families of AI as high-risk when used by, or on behalf of, competent public authorities (or EU bodies) in migration, asylum and border control:

• AI as a polygraph or similar tool — systems intended to detect the emotional state or truthfulness of a person.
• Risk assessment of persons entering — assessing a security, irregular-migration or health risk posed by a person who intends to enter, or has entered, the territory.
• Examination of applications — assisting competent authorities in examining applications for asylum, visa or residence permits, and the associated complaints, including assessing the reliability of evidence.
• Detection, recognition and identification of persons — in the context of migration, asylum and border control, with the explicit exception of verifying travel documents (a passport-gate 1:1 check is not, on its own, an Annex III point 7 system).

If your system does any of these things, it is high-risk. There is no "it's only decision-support" escape hatch here — Annex III point 7 is written to catch decision-support precisely because these systems shape life-altering outcomes.

---

Draw the prohibited line first

Before you scope a migration system as "high-risk to be documented," check it against Article 5 — because some uses in this space are not high-risk but banned outright, with no compliance route.

• Real-time remote biometric identification in publicly accessible spaces for law-enforcement purposes is prohibited under Article 5(1)(h), save for exhaustively listed and pre-authorised exceptions. A border zone that is a publicly accessible space, used for live biometric ID with a law-enforcement purpose, can fall on the prohibited side of this line — not the high-risk side.
• Risk assessment that predicts offending based solely on profiling of a person or their personality traits is prohibited under Article 5(1)(d). A migration "risk score" that crosses from assessing entry-risk into predicting criminality from profile alone leaves Annex III point 7 and becomes a prohibited practice.
• Untargeted scraping of facial images to build or expand recognition databases is prohibited under Article 5(1)(e).

The Article 5 prohibitions have been in force since 2 February 2025, and breaches sit in the top penalty tier: up to €35,000,000 or 7% of total worldwide annual turnover (Article 99(3)). Get this boundary right before anything else — a banned system documented as "high-risk" is still banned.

---

The duties that bite: FRIA, non-public registration, the full provider stack

Because point 7 systems are deployed by public bodies, two obligations stand out beyond the general high-risk requirements.

The Fundamental Rights Impact Assessment (Article 27). Public-body and public-service deployers must complete a FRIA before first use — assessing the system's impact on the fundamental rights of the people it affects. This is mandatory for migration and asylum deployers, not optional good practice. It is distinct from a GDPR Article 35 DPIA, though Article 27(4) lets a FRIA build on an existing DPIA rather than duplicate it.

Registration in the non-public section of the EU database. High-risk systems are registered in the EU database under Article 49 (the database itself is established by Article 71). For migration, asylum and border-control systems specifically, registration goes into a non-public, restricted-access section of that database — not the publicly searchable part — reflecting the sensitivity of the domain.

On top of those, the standard high-risk regime applies: a risk management system (Article 9), data governance (Article 10), technical documentation (Article 11 / Annex IV), human oversight (Article 14), accuracy, robustness and cybersecurity (Article 15), and — for the provider — conformity assessment under Article 43. Most Annex III points run the internal-control route (Annex VI); only point 1 biometrics requires a notified body.

---

The deadline caveat (write this down)

The deferral of the high-risk regime to 2 December 2027 has been adopted — the European Parliament voted on 16 June 2026 and the Council adopted it on 29 June 2026. Entry into force follows Official Journal publication, expected before 2 August 2026. Plan against the adopted 2 December 2027 deadline for stand-alone high-risk Annex III systems. The Article 5 prohibitions, by contrast, have already been in force and do not move.

---

Worked example: an immigration authority and its vendor

The Bundesamt für Migrationsdienste (BMD) — a fictional national immigration authority — procures a case-triage tool from VisaTriage OÜ, a fictional 40-person Estonian software vendor (~€6M turnover). The tool helps caseworkers prioritise asylum applications and flags inconsistencies in submitted evidence.

• VisaTriage OÜ is the provider (Article 16). The tool assists in examining asylum applications and assessing the reliability of evidence — squarely Annex III point 7. VisaTriage must run the Article 43 conformity assessment (Annex VI internal control), build the Article 11 / Annex IV technical file, the risk management system (Art 9), data governance (Art 10), and design in human oversight (Art 14). As an SME, its exposure is capped at the lower figure under Article 99(6).
• BMD is the deployer (Article 26). Before first use it must complete the Article 27 FRIA, register the system in the non-public section of the EU database (Art 49), assign trained human overseers, and ensure caseworkers do not defer automatically to the tool's flags.
• The line they must not cross. If a later release added a module that scored applicants' likelihood of future criminality from their profile alone, that module would leave Annex III point 7 and become prohibited under Article 5(1)(d) — no documentation can rescue it. The high-risk pathway only stays open while the system supports a human judgment grounded in objective, verifiable facts.

Same domain, two roles, one bright line — and both parties carry real, separate duties.

---

How Confir helps

Confir's classification logic tests a migration or border use against the Article 5 prohibitions first and only then against Annex III point 7, so a banned profiling or real-time-RBI use is flagged as prohibited rather than mis-scoped as "high-risk to be documented." For genuinely high-risk uses it assembles the provider stack, the Article 11 / Annex IV technical pack, and the Article 27 FRIA, and records the obligation to register in the non-public section under Article 49. The engine is deterministic and rule-based: the same intake produces the same finding, and every finding cites the Article or Annex point that fired.

Related guides

• Annex III point 7 — visa application AI
• Asylum application AI
• Border-control AI
• Article 5 prohibited practices
• The FRIA under Article 27
• How Annex III classifies high-risk AI