← 3D visual tours

AIGP study · Domain II

The AI Act risk tiers

Four tiers, one question: what does the system do to people? The higher the tier, the heavier the duties.

Source: Regulation (EU) 2024/1689, Articles 5, 6, 50 and 4. Study aid, not legal advice.

Scroll to move through the tiers. You can also use the numbered buttons on the right.

Tier 1 · Prohibited

Some uses are banned

Article 5 lists practices that are not allowed at all. One example: social scoring by public authorities that leads to unjustified or unrelated treatment of people (Article 5(1)(c)).

Prohibited practices sit outside the risk tiers: they are banned, not regulated.

Source: AI Act Art. 5(1)(c).

Tier 2 · High-risk

Heavy duties, in defined areas

Annex III lists areas such as employment, including recruitment and candidate ranking. Article 6 explains how a system is classed as high-risk.

Providers carry the main duties: risk management (Art. 9), data governance (Art. 10), technical documentation (Art. 11), logging (Art. 12) and human oversight (Art. 14).

Sources: AI Act Art. 6, Annex III, Art. 9, 10, 11, 12, 14.

Tier 3 · Transparency

Tell people it is AI

If a system talks directly with people, the Act expects those people to be told it is an AI, unless that is already plain from the context (Article 50(1)).

Source: AI Act Art. 50(1).

Tier 4 · Minimal risk

Few duties, not no duties

Minimal-risk systems have few mandatory duties. Article 4 still asks providers and deployers to take steps toward AI literacy, and voluntary codes of conduct are encouraged (Article 95).

Sources: AI Act Art. 4, Art. 95.

Why it matters

The stakes

Breaking the Article 5 bans can cost up to EUR 35 million. For a company, the 7% turnover figure applies instead if it is larger (Article 99(3)).

The Act does not switch on all at once, so check Article 113 before relying on any date.

Source: AI Act Art. 99(3), Art. 113.