The EU AI Act creates a tiered risk-based regime. The middle and most-regulated tier is "high-risk." Article 6 defines two routes into high-risk: (1) AI systems that are safety components of products covered by EU sectoral law listed in Annex I (e.g. medical devices, machinery, toys, vehicles), or (2) AI systems used in the eight areas listed in Annex III.
What are the eight Annex III areas?
Biometrics; critical infrastructure (water, gas, electricity, traffic); education and vocational training (admission, evaluation, monitoring); employment and worker management (recruitment, performance evaluation, task allocation); access to essential private services and essential public services (creditworthiness, public-benefit eligibility, emergency triage, life and health insurance); law enforcement (risk assessments, polygraph-like tools, evidence reliability); migration, asylum, and border control; administration of justice and democratic processes.
What does Article 6(3) "filter" do?
Article 6(3) added a filter: an AI system in an Annex III area is NOT high-risk if it does not pose a significant risk of harm to the health, safety, or fundamental rights of natural persons. Four conditions can lift it out of high-risk: (a) narrow procedural task, (b) improvement of human-completed activity, (c) detection of decision-making patterns or deviations, or (d) preparatory task to a relevant assessment. Profiling of natural persons always stays in high-risk.
What are the obligations for providers of high-risk systems?
A risk management system across the lifecycle (Article 9), data and data governance for training, validation, and testing (Article 10), technical documentation per Annex IV (Article 11), record-keeping/automatic logging (Article 12), transparency and instructions for use (Article 13), human oversight by design (Article 14), accuracy/robustness/cybersecurity (Article 15), a quality management system (Article 17), conformity assessment and CE-marking (Articles 43–49), registration in the EU database (Article 49), and post-market monitoring (Article 72).
What about deployers?
Deployers of high-risk AI systems have lighter but real obligations under Article 26: use the system per the provider’s instructions, ensure relevant input data is appropriate for the intended purpose, monitor operation and report incidents, retain logs, and — for certain Annex III systems used by public authorities or large private deployers — conduct a Fundamental Rights Impact Assessment (FRIA) under Article 27.
When do high-risk obligations apply?
Most high-risk obligations apply from 2 August 2026. For high-risk systems that are safety components of Annex I products (medical devices, machinery, etc.), the rules apply from 2 August 2027 to align with sectoral conformity-assessment cycles. Systems already on the market before 2 August 2026 only need to comply if they undergo "significant changes" after that date.