EU AI Act
In shortThe EU AI Act (Regulation 2024/1689) is the European Union's law that regulates AI systems by risk level, with duties that start in stages.
In plain words
The AI Act is a regulation of the European Union, published as Regulation (EU) 2024/1689. It follows a risk-based approach: the higher the risk of an AI system, the stricter the rules. The European Commission names four categories:
- Unacceptable risk: banned practices.
- High risk: allowed with strict duties, for example in areas such as hiring, education or biometrics.
- Transparency risk: duties to tell people when they deal with AI or see AI-made content.
- Minimal or no risk: no extra rules.
Providers of general-purpose AI models, the base of today's chatbots, have their own duties. Fines for serious breaches are high (Article 99 on EUR-Lex).
Why it matters
The rules start in stages. According to the Commission's AI Act page, checked on 26 October 2026:
| Date | What applies |
|---|---|
| 1 August 2024 | The Act enters into force |
| 2 February 2025 | Banned practices and AI literacy duties |
| 2 August 2025 | Rules for general-purpose AI models and governance |
| 2 August 2026 | General application, with some exceptions; transparency rules from August 2026 |
| 2 December 2027 | High-risk systems in listed areas (Annex III), extended from the first date |
| 2 August 2028 | High-risk systems in regulated products (Annex I), extended from the first date |
The later high-risk dates come from the AI Omnibus, a simplification package that the Commission page says has been in force since 27 July 2026. Dates have moved before, so check the Commission page and the official text.
For you as a user or a small firm:
- AI literacy: Firms that use AI should make sure staff know enough to use it responsibly.
- Transparency: Tell people when they deal with AI where the rules require it.
- High-risk use: Using AI to screen job applicants or to assess people is a high-risk area with strict duties.
- Outside the EU: The Act can also reach providers and users in other countries, such as Switzerland, when they place AI on the EU market or when the output of an AI system is used in the EU.
This is general information, not legal advice. Which duties apply to your firm depends on the facts.
Example
A Swiss firm uses a chatbot to answer customers, including some in Germany. It tells them they talk to an AI, trains the team and keeps personal data out of prompts (see what never to paste). It also wants an AI tool to rank job applicants. That is a high-risk area, so the firm checks the duties first and keeps a person in charge of each decision.
Common misconceptions
- "The AI Act bans chatbots." It does not. It bans specific practices and sets duties by risk level.
- "Only EU firms are affected." The Act has reach beyond the EU under conditions set in its text.
Questions and answers
When does the EU AI Act apply?
It entered into force on 1 August 2024. Banned practices apply from 2 February 2025, rules for general-purpose AI from 2 August 2025 and general application from 2 August 2026, with later dates for high-risk systems.
What are the risk categories of the AI Act?
Unacceptable risk, high risk, transparency risk, and minimal or no risk, according to the European Commission.
Does the EU AI Act apply to Swiss companies?
It can, if they place AI systems on the EU market or if the output of their AI is used in the EU. Check the text of the Act or get legal advice for your case.
Read next
- Large language model (LLM)
A large language model (LLM) is an AI system trained on huge amounts of text to predict the next token, which lets it write, summarise and answer.
- What not to paste into AI chatbots
Passwords, client data, IDs and secrets do not belong in a chatbot: a traffic-light checklist, provider data rules, Swiss and EU basics, and safer options.
- AI agent
An AI agent is a system where a model plans, uses tools and checks results in a loop to reach a goal with little step-by-step guidance.
- Open-weight model
An open-weight model is an AI model whose trained weights are published for download, so you can run and often adapt it yourself.
- How to choose an AI model
Match the model to the job: small for volume, mid for daily work, top or reasoning for hard problems and local for private data, with a decision tree and a test.