The EU's AI Act now requires labelling deepfake content

The persona

No single persona anchors this entry. The subject is Article 50 of the European Union's Artificial Intelligence Act, Regulation (EU) 2024/1689, setting transparency duties for the providers and deployers of certain AI systems rather than for any one company or character. The rule reaches any operator whose system serves users inside the EU, including a studio running a synthetic persona built on generative AI.

What the documents establish

The European Commission's own regulatory-framework page and its companion AI Act questions-and-answers page, both retrieved 16 September 2026, describe four duties under Article 50. Providers of AI systems interacting directly with people must design them so users know they are dealing with an AI system. Providers of generative AI systems must mark outputs in a machine-readable format, detectable as artificially generated, to the extent technically feasible. Deployers of emotion-recognition or biometric-categorisation systems must inform people exposed to them. And deployers of generative systems producing image, audio, or video content amounting to a deep fake must visibly disclose that the content is artificially generated, a duty extending to text published on matters of public interest. The Commission's FAQ states these obligations became applicable on 2 August 2026, a date that has already passed.

Character versus company

Article 50 does not ask what a synthetic persona says or claims about itself; it asks whether the entity that built or deployed it has attached a visible or machine-readable signal that a viewer is looking at AI-generated material. A persona's own posts or video output are the operator's material under this framework: it is the deployer, not the character, who bears the disclosure duty, and it is the provider of the underlying generative system, not the persona, who must build in machine-readable marking. The Commission's own text notes that specific exceptions apply to all of these obligations, though the retrieved pages describe the duties in general terms.

What to watch

The Commission's FAQ confirms the AI Office is still drafting a voluntary code of practice on transparency of AI-generated content, with more than two hundred stakeholders contributing, and that further guidelines on Article 50 are still forthcoming; the sources reviewed here show no finalized code or completed enforcement action. This is an editorial read: because the obligations became applicable only weeks before this entry was prepared, any claim that a specific operator is or is not compliant would need a dated, operator-specific check.

• Does a given persona's operator visibly disclose, in the EU, that its image, audio, or video output is AI-generated, as Article 50 requires of deployers?
• Has the AI Office published its guidelines or a finalized code of practice on Article 50 transparency obligations?
• Does the underlying generative-AI system a persona relies on mark its outputs in a machine-readable format, as required of providers?

Article 50 sets a disclosure duty on providers and deployers of AI systems, not a judgment on a persona's authenticity, and that duty has been technically applicable since 2 August 2026. Whether a given operator meets it is a fact to check against that operator's own practice, not the regulation's text alone.