EU AI Act

The EU AI Act (Regulation (EU) 2024/1689) is the world’s first comprehensive AI regulation. It entered into force in August 2024 and applies in stages. This entry provides factual orientation — it is not legal advice.

Risk-based approach

The AI Act assigns every application to one of four classes: unacceptable, high, limited and minimal risk. The higher the risk, the stricter the obligations. Prohibited practices such as social scoring are banned. What this means in operation is covered in Responsible AI & AI Act compliance.

What applies as of July 2026

Since 02.02.2025: the prohibitions and the AI literacy obligation for companies. Since 02.08.2025: obligations for GPAI models — transparency, copyright policy, training data summaries. From 02.08.2026, the EU AI Office gains full enforcement powers over GPAI providers.

What is being postponed

In May 2026, Council and Parliament reached a political agreement on the “Digital Omnibus”. It postpones the high-risk obligations: standalone systems (Annex III) to 02.12.2027, systems embedded in regulated products (Annex I) to 02.08.2028. Parliament approved the compromise on 16.06.2026 and the Council gave its final green light on 29.06.2026 — the regulation will be published in the Official Journal shortly. The GPAI obligations and the 02.08.2026 enforcement date are unaffected.

Why GPT, Claude & Co. are affected

ChatGPT, Claude and Gemini are built on GPAI models. The GPAI obligations target their providers — many have signed the GPAI Code of Practice. Companies as deployers mainly carry transparency and literacy obligations.

Last updated: July 2026