The EU AI Act made simple: what changes for you in 2026
What Regulation (EU) 2024/1689 is, its four risk levels, what’s banned, what must be labelled and the updated timeline after the “AI Omnibus”. Sources: European Commission, EUR-Lex and AESIA.
Informational content; not a substitute for professional advice.
The AI Act (Regulation (EU) 2024/1689) is, according to the European Commission, the first comprehensive legal framework on AI. It doesn’t regulate “AI” in the abstract, but specific uses according to their risk.
The four risk levels
| Level | Examples (European Commission) | What it means |
|---|---|---|
| Unacceptable | Social scoring, harmful manipulation, emotion recognition at work or school | Banned |
| High | CV sorting, exam scoring, access to credit | Strict obligations before going on the market |
| Transparency | Chatbots, generated content, deepfakes | Disclose and label |
| Minimal | Spam filters, video games | No new rules |
Updated timeline
- 2 February 2025: banned practices and the AI literacy obligation.
- 2 August 2025: rules for general-purpose models.
- 2 August 2026: general application and transparency rules.
- December 2026: new ban on AI “nudify” apps and generated child sexual abuse material.
- 2 December 2027: high risk in sensitive areas (Annex III), such as employment or education.
- 2 August 2028: high risk built into regulated products (Annex I).
The high-risk dates were pushed back by the so-called AI Omnibus, which entered into force on 27 July 2026.
What it means for you
- When you talk to a chatbot, you must be able to know it’s a machine.
- Deepfakes must be clearly labelled. Learn to spot them.
- If an AI decides about you in high-risk areas, human oversight will be mandatory once those rules apply.
Frequently asked questions
Does the AI Act ban using ChatGPT at work?
No. The Act regulates specific uses by risk. A general-purpose assistant isn’t banned; what matters is what it’s used for and your organisation’s policy.
Who enforces it in Spain?
The European Commission says that from 2 August 2026 the AI Office and national authorities apply and supervise the Act. In Spain, the Spanish AI Supervisory Agency (AESIA).
Glossary · Safety and good use
Keep reading
Sources
- Ley de IA (marco regulador) — Comisión Europea
- Reglamento (UE) 2024/1689 — EUR-Lex
- Agencia Española de Supervisión de la IA — AESIA
Change log
- : First published with the European Commission timeline after the AI Omnibus (in force on 27 July 2026).
Educational content, prepared with the help of AI tools, with sources cited. It is not a substitute for professional advice. Tools change fast: if you spot something out of date, write to us at contacto@cognenza.com. How we work.