Semantic Web Academy — home

EU AI Act: Tiers and What They Mean for You

Prohibited, high-risk, limited-risk, minimal — and the obligations attached.

0/1 done

Tiers + obligations

Four tiers

TierExamplesObligation summary
Unacceptable riskSocial scoring by governments, real-time biometric ID in public spaces.Banned.
High riskCredit, hiring, medical, education, critical infra, law enforcement.Risk mgmt, data governance, technical docs, transparency, human oversight, accuracy & robustness, conformity assessment, registration.
Limited riskChatbots, deepfakes.Disclosure that user is interacting with AI / content is AI-generated.
Minimal riskSpam filters, recommender systems (most).Voluntary codes of conduct.

Practical implications

  • Know your tier before you build.
  • For high-risk systems, the technical-documentation requirement is essentially: tracking + registry + model card + monitoring + audit log. (Sound familiar?)
  • Penalties scale up to 7% of global turnover for the worst breaches.

Analogy

Like food safety classifications: a salad bar and a shellfish counter both serve food, but the second has vastly stricter controls. The EU AI Act asks: which counter is your model behind?

Reading in progress · 0 of 1 activity done