AI Act: "AI-generated" labeling has been mandatory since August 2, what it means for content publishers

Since August 2, 2026, the transparency obligations of Article 50 of the AI Act apply across the European Union. They directly concern anyone publishing content produced with agents.

What's required

• Chatbots, agents and avatars interacting with the public must be clearly identified as such.
• Deepfakes and AI-generated informational text without human editorial review must be flagged.
• The European Commission has published icon templates ("AI", "AI-generated", "AI-modified") organizations can attach to their content.

Deadlines and penalties

The obligations apply from August 2 to new content; a grace period until December 2, 2026 is granted for AI systems already in service. A breach of Article 50 can bring a fine of up to €15 million or 3% of worldwide annual turnover, whichever is higher.

Why it matters for an agent-fed publication

The criterion for text is the absence of human editorial control. A site whose articles are drafted by agents and then reviewed and approved by a human isn't in the same position as one publishing on autopilot. In practice:

1. Document the editorial workflow: who reviews, when, with what record.
2. Clearly disclose AI use on the affected pages, with a notice and, where relevant, the official icon.
3. Identify conversational agents exposed to the public (forms, assistants, chatbots).
4. Have a lawyer validate the setup: this article summarizes trade press and isn't legal advice.

A side effect

OpenAI's dots and Meta's Muse aren't available in Europe for now. Always-on agents with account access raise GDPR and AI Act compliance questions that vendors will have to settle before opening these products here.

Sources: Blog du Modérateur, OIA, Droit & Technologies, Leto Legal, Possibility.