Date:

Share:

EU AI Act: What’s Now Mandatory for Chatbots and Generative AI

Related Articles

Since August 2, 2026, a new stage of the European Union's regulation on artificial intelligence, better known as the AI Act, has come into effect. Contrary to popular belief, this law isn't limited to large tech companies: it concretely changes how chatbots, virtual assistants, and content-generation tools must present themselves to the general public.

What's mandatory starting now

The “governance” component of the regulation is now fully operational. The European AI Office, in coordination with national authorities in each EU member state, officially holds its oversight powers and can initiate enforcement procedures. In practical terms, two transparency obligations now apply to providers of AI systems:

  • Chatbots and conversational assistants must disclose at the start of an interaction that the user is dealing with an artificial intelligence, unless this is already obvious to a reasonably well-informed user.
  • Content generated or altered by AI — images, audio, manipulated videos (deepfakes), synthetic text — must be identifiable as such by the people who view or hear it.

Tighter oversight of large models

The European Commission has also gained expanded powers over providers of general-purpose AI models, the category that includes today's most widely used large conversational agents. It can now request additional information from these providers and impose penalties in cases of confirmed failure to meet transparency or technical documentation requirements.

For a French company that integrates a chatbot on its website or uses a content-generation tool in its communications, the practical consequence is simple: it's now necessary to verify that these tools clearly disclose their automated nature, or risk exposing the service's provider to a report to the relevant authorities.

The “high-risk” component isn't in effect yet

Another frequently misunderstood point: the regulation's heaviest obligations, those concerning so-called “high-risk” AI systems (automated recruitment, credit scoring, certain uses in healthcare or education), don't apply yet. A delay was proposed by the European Commission in late 2025, and in spring 2026 the Council and the European Parliament agreed to simplify the timeline: these requirements, originally set for August 2, 2026, are now expected by the end of 2027. Affected companies therefore have extra time to adapt their systems, though this doesn't exempt them from the transparency obligations already in force.

This two-stage timeline illustrates the challenge facing European regulators: governing a technology that evolves faster than the legislative process, without slowing down a sector where competitiveness plays out on a global scale. For individuals, the key takeaway remains simple: from now on, any site or service that uses AI to interact with you must, in principle, tell you so.

For more on companies' regulatory obligations, Pitas.com's Law and Justice section tracks the main developments to watch. Digital professionals can also check out the High Tech – Digital section, while business leaders will find practical pointers in the Business – Enterprise section.

Frequently asked questions

Does the EU AI Act apply to all companies?

Yes. As soon as a company offers a service using artificial intelligence to users located in the European Union, it is subject to the transparency obligations that took effect on August 2, 2026, regardless of its size.

What happens to a company that doesn't comply with these new rules?

National authorities, coordinated by the European AI Office, can request additional information from the providers concerned and initiate enforcement procedures in cases of confirmed failure to meet transparency obligations.

Are high-risk AI systems already regulated?

No: these stricter obligations have been postponed to the end of 2027, following an agreement between the Council and the European Parliament aimed at simplifying the regulation's original timeline.

Sources

This article was written with the help of artificial intelligence. Editorial policy

Popular Articles

Editorial policy and use of artificial intelligence