France's SREN law offers a defence for labelling algorithmic content

The persona

This entry has no single named persona. It reads France's enacted SREN law for what it actually requires of anyone who shares an algorithmically generated image, voice or likeness of a real person, a question that reaches any virtual-persona content built from a real individual's face or voice for a French audience. LOI n° 2024-449, the law "visant à sécuriser et réguler l'espace numérique," was promulgated on 21 May 2024 and published the following day in the Journal Officiel. The source text is in French; the description below reflects the enacted articles as retrieved on 16 September 2026, and readers should not extend it beyond what those articles state.

What the documents establish

Article 15 of the enacted law amends Article 226-8 of the Penal Code, which already punished publishing a non-consensual "montage" using a person's image or words. The amendment extends the offence to cover a visual or audio "content généré par un traitement algorithmique" (content generated by an algorithmic process) that represents a person's image or words without their consent. Critically, the text builds in a defence: the offence does not apply if it is obviously algorithmic, "or if it is not expressly identified as such." Sharing such content via an online public communication service raises the penalty to two years' imprisonment and a €45,000 fine, per the Journal Officiel publication of the same law. Article 21 separately creates Penal Code Article 226-8-1, targeting non-consensual sexual "montage" and algorithmically generated sexual content, with a higher penalty of three years and €75,000 when distributed online.

Character versus company

The statute does not create a general duty to label synthetic content; it withholds a criminal-law defence from anyone who fails to disclose that a depiction of a real, identifiable person is algorithmically generated. That is a narrower, more targeted rule than a blanket "label your AI content" mandate, and it applies to the depicted individual's likeness, not to a fictional persona's own invented backstory. An operator publishing a persona whose look or voice draws on a real person's identity is the party this law reaches; the persona's own posts carry no separate legal status.

What to watch

• Has a French court applied Article 226-8 or 226-8-1 to a commercial virtual-persona campaign rather than to non-consensual imagery generally?
• What counts, in practice, as content that "obviously" appears algorithmic without an explicit label?
• How does this criminal-law defence interact with the EU AI Act's separate labelling duties once both apply in France?

Treat this as one criminal-law provision among several instruments now converging on synthetic media in France, not as the country's complete regulatory answer to virtual personas.