EU AI Act for Public Authorities: Law Enforcement, Migration, Border Control and Justice AI Under Annex III(6)-(8)
When a bank gets AI wrong, a customer complains. When a police force, border agency or court gets it wrong, someone can lose liberty, a visa or a legal claim. That is why Annex III of the EU AI Act dedicates three whole points (6, 7 and 8) to law enforcement, migration and border control, and the administration of justice and democratic processes. This guide explains what is covered, what extra rules apply to public bodies, and when. For the outright bans, see our Article 5 guide.
Annex III point 6: law enforcement
High-risk uses, when intended for or on behalf of law enforcement authorities or EU bodies supporting them, include:
- assessing the risk of a person becoming a victim of a criminal offence;
- polygraphs and similar tools;
- evaluating the reliability of evidence in investigations or prosecutions;
- assessing the risk of a person offending or re-offending, where not based solely on profiling, or assessing personality traits and past behaviour;
- profiling of persons in the course of detection, investigation or prosecution.
Note the interaction with Article 5(1)(d): predicting that an individual will commit a crime solely on the basis of profiling or personality traits is banned, not merely high-risk. Real-time remote biometric identification in public spaces for law enforcement is likewise largely prohibited under Article 5(1)(h); see our biometrics guide.
Annex III point 7: migration, asylum and border control
High-risk uses, when intended for or on behalf of competent public authorities, include:
- polygraphs and similar tools;
- assessing security, irregular migration or health risks posed by a person entering or seeking to enter a Member State;
- assisting in the examination of applications for asylum, visas and residence permits, and related eligibility and evidence assessments;
- detecting, recognising or identifying people in the migration context, with the exception of verifying travel documents.
Civil society has pressed for stricter limits in this area; see for example the joint statement from Privacy International and others. The final Act keeps these systems in the high-risk tier rather than banning them.
Annex III point 8: justice and democratic processes
Point 8 covers two uses:
- AI intended to be used by or on behalf of a judicial authority to research and interpret facts and the law and to apply the law to facts, or used similarly in alternative dispute resolution;
- AI intended to influence the outcome of an election or referendum, or the voting behaviour of people exercising their vote. The category excludes tools whose outputs people are not directly exposed to, such as campaign administration and logistics.
What is different for public bodies
| Duty | What it means | Reference |
|---|---|---|
| Fundamental rights impact assessment | Bodies governed by public law must assess impact on rights before first use of a high-risk system | Art. 27 |
| EU database registration | Public authorities deploying high-risk systems register their use | Art. 49(3) |
| Non-public registration | Law enforcement, migration, asylum and border control systems are registered in a secure non-public section | Art. 49(4) |
| Deployer duties | Instructions, human oversight, input data quality, monitoring, log retention, informing affected people | Art. 26 |
| Supervision | For these systems the market surveillance authority is generally the data protection authority | Art. 74(8) |
Check each reference against the official text on EUR-Lex before relying on it; the Act's registration and supervision rules for these sectors are detailed and have exceptions. Our guides to Article 26, Article 27 FRIA and enforcement architecture cover the mechanics.
Buying AI as a public authority
Most authorities procure rather than build. Under Article 25, repurposing a purchased system for a high-risk purpose can turn the buyer into a provider. Use the evidence pack in our procurement guide: intended-purpose statement, instructions for use, logging capability, conformity evidence, and the vendor's own classification reasoning.
Timeline
The Article 5 bans and Article 4 AI literacy duty have applied since 2 February 2025. For the Annex III high-risk obligations covered here, the Digital Omnibus (Regulation (EU) 2026/1744) moved the application date to December 2027; see our implementation status post for the full re-baselined timeline. The delay is not a reason to wait: logging, oversight design and vendor contracts take many months, and the notified body bottleneck shows how infrastructure can lag.
Readiness checklist for authorities
- Inventory every AI-enabled tool used in policing, border, asylum, court and electoral work, including tools embedded in case-management software.
- Classify each against Annex III points 6, 7 and 8 and Article 5.
- Stop or redesign anything that fits a prohibition now; the bans are already enforceable.
- Assign trained human overseers (Article 14 and Article 4) and document their authority to override.
- Draft the FRIA template and registration records.
- Update tender templates and contracts with the AI Act evidence pack.
Does the AI Act apply to police use of AI?
Yes. Annex III point 6 lists law enforcement uses as high-risk, and Article 5 prohibits some, such as individual predictive policing based solely on profiling.
Is AI used to process asylum applications high-risk?
AI intended to assist in examining asylum, visa and residence applications is listed in Annex III point 7.
Is a court's use of an AI legal research tool high-risk?
Annex III point 8 covers AI used by or for judicial authorities to research and interpret facts and law and apply the law to facts.
When do these high-risk obligations apply?
After the Digital Omnibus, Annex III obligations apply from December 2027, while the Article 5 bans have applied since 2 February 2025.
This article is general information, not legal advice.
Related reading
Article 14 EU AI Act: Engineering Human Oversight That Actually Works
Article 14 is a design obligation, not an HR policy. Here's how to engineer the five oversight capabilities, split provider vs deployer duties, handle the biometric two-person rule, and build an audit-ready evidence trail.
EU AI Act Article 13: What Your Instructions for Use Must Contain, and What Deployers Should Demand From Vendors
Article 13 turns the instructions for use into the contract between a high-risk AI provider and its deployers. Here is what Article 13(3) requires, the evidence deployers should ask for, and where harmonised standards stand after 14 October 2026.
GPAI Models With Systemic Risk: The Safety and Security Chapter of the Code of Practice, Explained
Providers of the most capable general-purpose AI models carry extra duties. Here is what the Safety and Security chapter of the GPAI Code of Practice asks for, the incident timelines, and the dates that matter now the AI Office can enforce.