The EU AI Act lists specific recruitment and employment-management uses as high-risk AI under Annex III. An exception under Article 6(3) requires both a qualifying task and no significant risk of harm; systems that profile natural persons remain high-risk. The European Commission says specified transparency provisions have applied since August 2, 2026, while the Annex III high-risk rules apply from December 2, 2027. The Commission’s Annex III categories, classification rules and enforcement timeline set the scope and dates.
For employers, classification turns on what the AI system is intended to do and how it fits into a decision—not simply on whether a vendor or HR team calls it administrative. These are EU AI Act requirements.
Classify each system by its intended use and influence on decisions, then track deadlines and duties against the specific rule that applies.
Which HR AI uses are covered by Annex III?
Annex III covers AI intended for recruitment or selection, including targeted job advertising, application analysis and filtering, and candidate evaluation. It also lists systems intended to influence decisions about work relationships, promotion or termination; allocate tasks based on a person’s behavior or traits; or monitor and evaluate workers’ performance or behavior. The Commission’s Annex III text sets out these use cases.
A practical first step is to map what each system is intended to do in the hiring or employment process. A tool that filters applications, for example, falls within a listed recruitment use; a system used to allocate work based on individual behavior or traits falls within a listed employment-management use.
When can an Annex III system qualify for the exception?
An Annex III system qualifies for the Article 6(3) exception only if it poses no significant risk of harm—including by not materially influencing a decision—and meets at least one of the provision’s task conditions. Those conditions cover narrow procedural tasks, improving a human activity already completed, certain pattern-detection tasks, and preparatory tasks. Article 6(3) also says an Annex III system that profiles natural persons is always high-risk.
That means the label “procedural” or “preparatory” does not settle the classification: the risk threshold must also be met. Employers assessing a particular system need to consider its intended purpose, its influence on decisions, whether it profiles people and how its task fits the listed conditions.
Which dates apply to each requirement?
The Commission’s timeline distinguishes specified transparency provisions from Annex III high-risk rules and from rules for high-risk AI embedded in regulated products. The dates apply to different obligations, not one shared start date. The Commission’s enforcement timeline gives these milestones:
| Requirement | Date |
| Specified Article 50 transparency provisions apply | August 2, 2026 |
| Deadline for providers of systems placed on the market before August 2, 2026, for the specified Article 50(2) marking and detection obligation | December 2, 2026 |
| Annex III high-risk rules apply | December 2, 2027 |
| Rules for high-risk AI embedded in regulated products apply | August 2, 2028 |
The December 2, 2026 deadline concerns providers of systems already on the market before August 2, 2026, and the specified Article 50(2) marking and detection obligation. The December 2027 date is the one for Annex III high-risk rules.
What must an employer deploying high-risk AI do?
Article 26 sets out responsibilities for employers deploying high-risk systems. Employers must use a system according to its instructions, assign human oversight to people with the necessary competence, training and authority, monitor the system, and ensure that input data under their control are relevant and sufficiently representative. The Commission’s Article 26 text also requires employers to inform workers’ representatives and affected workers before workplace use, under applicable information procedures. For in-scope Annex III systems that make or assist decisions about natural persons, the people subject to those decisions must be informed as well.
If an employer has reason to believe that using a high-risk system may create a risk, it must suspend use and inform the provider or distributor and the relevant market-surveillance authority without undue delay. Serious incidents trigger a separate immediate reporting duty.
The log-retention rule is qualified: employers retain automatically generated logs only to the extent those logs are under their control. Retention must suit the purpose and last at least six months, unless applicable EU or national law provides otherwise. Six months is therefore a minimum for covered logs, not a fixed deletion date for every log.
A useful compliance sequence is to map intended uses, assess the full Article 6(3) test, match each obligation to its date, and assign the relevant oversight, information, monitoring and recordkeeping responsibilities.