The EU AI Act: what changes for AI-generated images

The European regulation on artificial intelligence (the AI Act), adopted in March 2024 with its first provisions taking effect in February 2025, is the world's first comprehensive legal framework governing AI. Among its most concrete measures: the obligation to label content generated by artificial intelligence, including images. For companies, platforms and individuals, these new rules profoundly change how AI images can be used.

What does the AI Act say about AI-generated images?

Article 50 of the AI Act imposes a transparency obligation for AI-generated content. In practice, anyone who uses an AI system to generate or manipulate an image, audio or video constituting a deepfake must disclose that the content has been artificially generated or manipulated. This obligation applies to AI system providers (companies like OpenAI, Midjourney, Stability AI), who must build labeling mechanisms into their tools, but also to users of these systems, who must disclose the artificial nature of the content when they publish it publicly. The marking must be machine-readable and, as far as possible, detectable by humans. The recommended technical standard is C2PA (Content Credentials), which embeds provenance metadata directly into the image file.

What penalties are provided for?

The AI Act provides for graduated penalties. Breaches of the transparency obligations can result in fines of up to 15 million euros or 3% of global annual turnover for companies. For providers of general-purpose AI models (like image generators), penalties can reach 35 million euros or 7% of global turnover. These amounts are comparable to GDPR fines and reflect Europe's determination to enforce these rules. The national authorities of each member state are responsible for oversight and enforcement, coordinated by a European AI Office based in Brussels.

What is the rollout timeline?

The AI Act applies gradually. Bans on unacceptable-risk AI systems (subliminal manipulation, social scoring) came into force in February 2025. The transparency obligations for AI-generated content, including the labeling of images, apply from August 2025. The full rules for high-risk AI systems will come into force in August 2026. As of March 2026, we are therefore in the period when the AI image labeling obligations are fully applicable, and the first oversight and enforcement procedures are expected in the coming months.

What impact for marketplaces and e-commerce?

For online sales platforms, the AI Act creates a new regulatory context. Product images generated or retouched by AI must now be identified as such. Marketplaces that allow listings with AI-generated photos without appropriate labeling expose themselves to penalties. This requirement reinforces the value of photo-certification mechanisms like CertiPix: a photo certified at the moment of capture proves not only that it is authentic, but also that it doesn't fall under the AI labeling obligations, since it comes from a real physical sensor. For sellers, requiring certified photos becomes an argument for trust and regulatory compliance, not just an anti-fraud tool.

How is France transposing the AI Act?

France, through the CNIL and the future national AI authority, is preparing to transpose the AI Act's obligations into national law. The CNIL has published several recommendations on the use of AI within the GDPR framework, and its role naturally extends to overseeing the transparency obligations on AI content. For sole traders and SMEs, the key is to understand that these obligations apply to everyone, not just to big tech companies. If you use an AI image generator to create product photos, marketing visuals or illustrations, you are subject to the labeling obligation from August 2025.

With the AI Act, the European Union establishes a clear principle: any AI-generated image must be identified as such. Photo certification at the source becomes a compliance tool, not just a security one.