Synthetic Media on Labelling of AI-Generated Content: EU Rules

Synthetic media on labelling of AI-generated content is now a legal duty under the EU AI Act. Here’s what creators and deployers must do, with clear steps and…

· 5 min read

What Synthetic Media and AI-Generated Content Labelling Means Today

Synthetic media on labelling of AI-generated content is a legal requirement for many creators and businesses under the EU AI Act. From 2 August 2026, providers of generative AI systems must mark synthetic audio, image, video and text output in a machine-readable format so it can be detected as artificially generated or manipulated. This applies to any content produced by AI and published or shared after that date. Content created and published before 2 August 2026 does not need retroactive labelling.

The marking must be effective, interoperable, robust and reliable as far as technically feasible. In practice, this means using a combination of metadata (such as C2PA Content Credentials) and invisible watermarks embedded in the file itself. Both help ensure detection even if the content is edited, shared across platforms or stripped of metadata. The EU has also published voluntary icons to signal whether content is fully AI-generated, partially modified or involves AI in another way.

Who Must Label Synthetic Media and AI Output Under the EU AI Act

The EU AI Act’s Article 50 transparency obligations apply to providers and deployers of AI systems. Providers of generative AI must mark synthetic audio, image, video and text output so it can be detected as AI-generated or manipulated. Deployers must disclose deepfakes and AI-generated or manipulated text published to inform the public on matters of public interest. Deployers of emotion recognition or biometric categorisation systems must also inform people exposed to them.

There are narrow exemptions. Outputs used in business-to-business or industrial contexts are exempt from the marking duty. Artistic, creative, satirical or fictional works only need disclosure in a way that does not hamper the display or enjoyment of the work. AI-generated text under human review with editorial responsibility is exempt from the public-interest text disclosure duty. These exemptions do not apply to other duties, such as the requirement to inform people when they interact with an AI system.

How to Label Synthetic Media and AI Content: Practical Steps

To comply with the marking duty, use a multi-layered approach. First, embed machine-readable metadata such as a C2PA manifest, which is cryptographically signed to bind claims about origin and edits to the file. Second, add an invisible watermark directly into the content, as metadata can be stripped by screenshots or platform processing. Third, consider using the EU’s voluntary icons to signal AI involvement to human audiences.

For deployers, the disclosure must be clear, distinguishable and provided at the latest at first interaction or exposure. For example, a news outlet publishing AI-assisted text on a matter of public interest must label it as such at the top of the article. For deepfakes, the disclosure must be similarly prominent. The EU’s Code of Practice on Transparency of AI-Generated Content, finalised in June 2026, provides a voluntary framework for providers and deployers to demonstrate compliance with these duties.

What Counts as Synthetic Media and AI-Generated Content

Synthetic media and AI-generated content include any audio, image, video or text created or manipulated by an AI system. This ranges from fully AI-generated images (such as those produced by text-to-image models) to AI-edited videos (such as deepfakes) and AI-assisted text (such as articles written with generative AI tools). The EU AI Act’s marking duty applies to all these formats, provided they are published or shared after 2 August 2026.

Examples of synthetic media include AI-generated music, voice clones, AI-upgraded photos and AI-summarised news articles. The duty to mark applies regardless of whether the content is used commercially, for entertainment or for information, unless an exemption applies. For instance, a studio using AI to generate concept art for internal use (B2B) does not need to mark it, but the same image shared publicly would require labelling.

Tools and Standards for Synthetic Media Labelling

C2PA (Coalition for Content Provenance and Authenticity) is an open technical standard that enables creators to attach cryptographically signed metadata, known as Content Credentials, to their files. This metadata can include information about the origin of the content, edits made and the tools used. However, metadata can be removed when files are shared or processed, so it should be combined with invisible watermarks embedded in the content itself.

The EU’s Code of Practice on Transparency of AI-Generated Content recommends a multi-layered approach, including metadata and watermarks, to ensure robustness. Providers and deployers can also use the EU’s voluntary icons to signal AI involvement to human audiences. Some tools, such as L'AItelier DNA, certify files with tamper-evident provenance certificates, combining C2PA manifests for images and video with invisible watermarks for audio and documents.

Deadlines and Grace Periods for Synthetic Media Labelling

The Article 50 transparency obligations under the EU AI Act apply from 2 August 2026. However, the Digital Omnibus on AI, published in July 2026, gives generative AI systems placed on the market before 2 August 2026 until 2 December 2026 to meet the marking duty. This grace period does not apply to other duties under Article 50, such as informing people when they interact with an AI system or disclosing deepfakes.

Content generated and published before 2 August 2026 does not need to be labelled retroactively. The Commission’s non-binding Guidelines on Article 50, published in July 2026, provide further clarification on how to meet the marking duty, but they do not change the legal requirements.

Penalties for Failing to Label Synthetic Media

Breaching the Article 50 obligations can lead to fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. Lower ceilings apply to SMEs. These penalties are designed to ensure compliance with the transparency duties, which are critical for public trust in AI-generated content. Deployers and providers should ensure they have processes in place to meet the marking, disclosure and notification requirements.

Frequently asked questions

What is synthetic media in the context of AI?

Synthetic media refers to any audio, image, video or text created or manipulated by an AI system. This includes fully AI-generated content, such as images or music, as well as AI-edited or enhanced content, like deepfakes or AI-assisted articles.

Do I need to label AI-generated content created before August 2026?

No. Content generated and published before 2 August 2026 does not need to be labelled retroactively under the EU AI Act.

What are the exemptions for labelling synthetic media?

Exemptions include outputs used in business-to-business or industrial contexts, artistic or fictional works where disclosure would hamper enjoyment, and AI-generated text under human review with editorial responsibility for public-interest matters.

How can I check if a file has been labelled as AI-generated?

You can use tools that verify C2PA metadata or invisible watermarks. For example, the public validator at laitelier.com/validator allows anyone to check a file’s provenance certificate.

What is the difference between metadata and watermarks for labelling?

Metadata, such as C2PA Content Credentials, is attached to the file and can be stripped by processing or sharing. Invisible watermarks are embedded directly into the content, making them more robust against removal.

Put provenance on your next file

L'AItelier signs images and video with C2PA, adds an invisible watermark, and issues a certificate anyone can check. Your file is deleted as soon as the certificate is issued.

General information about the law as it stood when this was published, not legal advice for your situation.