The AI Act in your shop: transparency duties since August 2026

01The timeline in one minute

• February 2025: prohibited practices and the AI literacy duty apply.
• August 2025: duties for providers of large AI models.
• August 2026: transparency duties and large parts of the remaining framework.
• Individual high-risk duties follow staged until 2027.

Brussels keeps debating delays for individual parts. Do not count on it: the transparency duties for customer contact apply.

02What transparency means in practice

• A chatbot must identify itself as AI, unless it is obvious.
• AI-generated or heavily edited images and videos that look real must be labelled.
• Providers of AI systems must mark content machine-readably. As the user you take care of the visible label.
• Emotion recognition and biometric categorisation require informing the people affected, rarely relevant in retail.

03Am I high-risk?

For typical shops the answer is almost always no. Product recommendations, text generation and support bots are not high-risk. It gets critical with topics like credit scoring or applicant screening done by AI. Then much harder duties apply.

If you only use ready-made tools, the law considers you a deployer. That means fewer duties than the provider of the system, but not zero.

04The underrated duty: AI literacy

Since February 2025, Article 4 requires that everyone working with AI in your business has sufficient AI literacy. A short, documented training session covers many cases. No training and no documentation is an unnecessary risk.

By the way: this is exactly why every post in this wiki says that AI helped create it. Transparency costs nothing and builds trust.