High-risk AI systems
The eight Annex III areas
- Biometrics, including remote identification and emotion recognition
- Critical infrastructure — traffic, water, gas, heating, electricity
- Education and vocational training — admission, assessment, proctoring
- Employment and worker management — recruitment, promotion, task allocation, termination
- Access to essential public and private services, including credit scoring
- Law enforcement
- Migration, asylum and border control
- Administration of justice and democratic processes
The category most companies meet without looking for it is employment and worker management: software that screens applications, ranks candidates, allocates shifts or scores performance. See also employee data protection.
When Annex III does not mean high-risk
Article 6(3) allows a system that sits within an Annex III area to be treated as not high-risk, where it performs only a narrow procedural task, improves the result of a previously completed human activity, detects decision patterns without replacing the human assessment, or performs a preparatory task.
The exemption never applies where the system performs profiling of natural persons. And relying on it is not a decision taken in passing: the assessment has to be documented before the system is used.
What is required
Risk management system
A continuous process across the whole lifecycle: identifying foreseeable risks, estimating them and adopting mitigation measures.
Data governance
Training, validation and testing data sets have to be relevant, representative and, as far as possible, free of errors and bias.
Technical documentation
Drawn up before the system is placed on the market and kept up to date, demonstrating that the requirements are met.
Record keeping
Automatic logging of events over the lifetime of the system, so that its operation can be traced afterwards.
Transparency
Operation sufficiently transparent that deployers can interpret the output and use it appropriately, with instructions for use.
Human oversight
Designed so that natural persons can oversee it effectively — including the ability to decide not to use the output, or to intervene and stop it.
Accuracy, robustness and cybersecurity
An appropriate level of each, consistent throughout the lifecycle, with resilience against attempts to alter the system’s behaviour.
The deadline, and why it is not far off
Under Regulation (EU) 2026/1744, the Digital Omnibus on AI, in force since 27 July 2026, compliance for Annex III high-risk systems is required by 2 December 2027.
The inventory and the classification do not depend on that date and are done once. The documentation, the oversight arrangements and the data governance take considerably longer, and they cannot start until the inventory exists.
Frequently asked
How do we know whether a system is high-risk?
Check whether the use falls into one of the eight categories in Annex III: biometrics, critical infrastructure, education, employment and worker management, access to essential public and private services, law enforcement, migration and asylum, or the administration of justice. The European Commission guidelines of May 2026 give concrete examples.
What is the deadline?
Regulation (EU) 2026/1744 (the Digital Omnibus on AI), in force since 27 July 2026, moved the compliance deadline for high-risk systems under Annex III to 2 December 2027. Systems embedded in products regulated under Annex I follow from 2 August 2028.
We only use such a system, we did not build it. Does this apply to us?
Yes, in part. The provider carries the heavier set of obligations, but the deployer has its own: ensuring human oversight, using the system in line with the instructions, monitoring its operation and keeping the logs. Those duties cannot be contracted away to the supplier.
Does an exemption exist?
Article 6(3) allows a system that falls within an Annex III area to be treated as not high-risk where it performs only a narrow procedural task or improves the result of a previously completed human activity. The exemption never applies where the system performs profiling of natural persons. The assessment has to be documented before relying on it.
Do any of your systems fall under Annex III?
The initial audit establishes which systems you use and how each is classified.
Request a free audit