EU AI Act
Regulation (EU) 2024/1689 — Artificial Intelligence Act, as amended by Regulation (EU) 2026/1744
Topic : Governance · Level : Regional
Today
Prohibitions, GPAI and Art. 50 transparency apply. Digital Omnibus (2026/1744) delays high-risk rules to Dec 2027 and Aug 2028.
Summary
The AI Act regulates AI by risk level. Bans and general-purpose AI rules already apply. The 2026 Digital Omnibus postponed high-risk obligations to December 2027 (standalone) and August 2028 (in products). Fines reach 7% of turnover.
Who is concerned
- Turnover
- non précisé
- Employees
- non précisé
- Listed company
- cotées et non cotées
- Sectors
- AI providers
- Machinery and devices
- HR and credit scoring
- All AI deployers
- Triggering activities
- Placing an AI system on the EU market
- Using AI output in the EU
- Embedding AI in regulated products
- Extraterritorial reach
- Yes
Applies to non-EU providers whose AI or its output is used in the EU; EU authorised representative required.
- Examples
- Chinese machine maker with AI safety part
- SaaS vendor scoring EU job applicants
What the law requires
- Classify each AI system by risk level
- Meet high-risk duties: risk management, data, logs, oversight
- Label AI-generated content and disclose chatbots (Art. 50)
- Take measures to support staff AI literacy
Penalties
Up to €35M/7% (prohibited AI), €15M/3% (other duties), €7.5M/1% (false info). SMEs: lower amount applies; small mid-caps too, except bans.
What this changes for a supplier in China
Chinese makers of machinery, devices or software with AI functions sold in the EU must document conformity and appoint an EU representative; the delay gives until 2027-2028.
Expected evidence
- AI system inventory and risk classification
- Technical documentation and logs
- EU declaration of conformity
Tiers : Tier 1