|
|
The EU AI Act entered into force on 1 August 2024 and applies in stages. Here are the key dates and what they mean in practice, as orientation from FHC+P's implementation practice, not legal advice.
Prohibited AI practices (e.g. social scoring, certain biometric applications) may no longer be used. At the same time Article 4 applies: whoever uses AI must ensure sufficient AI literacy among the employees involved.
Should be done: AI inventory (including shadow AI), check against the prohibition list, documented basic training for all employees who work with AI.
Obligations for providers of general-purpose AI models (transparency, copyright, technical documentation), plus governance structures and member-state penalty rules.
Should be done: For pure users mainly relevant with own AI products built on third-party models: know and document the origin and contract terms of the models used.
The regulation now applies in large parts: transparency and labelling obligations (Article 50) for chatbots, AI-generated content, and deepfakes, as well as the requirements for high-risk systems under Annex III. AI content must be recognisable or marked machine-readably, with exemptions for editorially reviewed content under human responsibility.
Should be running now: Labelling and approval processes for AI content (with proof), chatbot transparency, risk classification of all AI use cases, and, where high-risk systems are used, the associated deployer obligations. Our Article 50 self-check shows where you stand in 10 questions.
Transition periods end: among others for high-risk AI as part of regulated products (Annex I) and for general-purpose AI models that were already on the market before August 2025.
Should be planned: A roadmap for affected legacy systems and products, so the final stage does not become a project under time pressure.
With an AI inventory you will know within days which deadlines affect you and what needs to be done. More on our page Implementing the EU AI Act.
All application stages with date and duty as a fillable table: applies to us?, status, owner, plus the to-dos per stage. We send you the download link by e-mail.
Have further questions? Feel free to contact us directly via our contact form or by e-mail.
2 August 2026: since then the transparency and labelling obligations of Article 50 apply for chatbots and AI-generated content, as well as the requirements for high-risk systems under Annex III. Any company using AI in marketing, customer contact, or decision processes is practically always affected.
Yes. The obligations depend on the risk of the use case, not on company size. Even a five-person business that operates a chatbot or publishes AI content is subject to the transparency obligations and must ensure AI literacy.
The regulation provides for substantial fines, scaled by the type of violation and company revenue. In day-to-day business, however, the reputational and liability aspect matters more than the threat of fines: whoever documents and labels AI use cleanly can also give clear answers to customers and clients.