How it works
Obligations depend on risk. The European Commission often sums up the text in four levels. It's a reading aid: the regulation itself doesn't name "limited risk" or "minimal risk" categories.
- Prohibited practices: some uses of AI are banned.
- High risk: heavy obligations apply to the cases the regulation lists, such as risk management, technical documentation and human oversight.
- Transparency: people must know they're dealing with an AI, and providers must mark generated content in a machine-detectable way (Article 50).
- Other systems have no obligation tied to their risk level. All providers and deployers must still train their staff on AI (Article 4).
The text also sets rules for general-purpose AI models, the kind many assistants are built on. And it separates the provider, who develops the system, from the deployer, who uses it in their business. The timeline:
- 1 August 2024: entry into force;
- 2 February 2025: prohibited practices and staff AI literacy (Article 4);
- 2 August 2026: general application, including the duty to tell people they're interacting with an AI (Article 50);
- 2 December 2026: marking of generated content for systems placed on the market before 2 August 2026, a deadline added by Regulation (EU) 2026/1744;
- 2 December 2027: obligations for the high-risk systems listed in Annex III, postponed by that same 2026 regulation.
Why it matters in e-commerce
A brand that puts an AI assistant on its site becomes its deployer. At a minimum, it has to tell shoppers they're talking to an AI, and make sure its provider documents its own compliance. Breaching the Article 50 transparency obligations can cost up to €15 million or 3% of worldwide annual turnover (Article 99).
For a team building its own assistant in-house, documenting GDPR and EU AI Act compliance is part of the job, alongside keeping the catalog in sync and protecting the assistant against misuse.
Real-life example: iadvize.com
On iadvize.com, the terms of use state that shoppers interacting with the AI Shopping Assistant are informed of it, as required by Article 50 of the AI Act, and that AI-generated content is labeled. The privacy policy places the AI systems iAdvize uses in the "limited risk" category, the one for systems subject to the Article 50 transparency obligations.
Common confusion
The EU AI Act and GDPR often get mixed up. GDPR protects personal data, whatever tool processes it. The AI Act governs AI systems themselves: their risk level, their transparency, their documentation. An AI assistant that processes shopper data falls under both.