Supporting the European Code of Practice on AI-generated content

Today we're confirming Synthesia's support for the final version of the European Commission's Code of Practice on marking and labelling of AI-generated content. The Code is the voluntary instrument the AI Office has built to help providers and deployers of generative AI systems meet their transparency obligations under Article 50 of the AI Act, which takes effect in August 2026.

We were among the first AI companies to join the Coalition for Content Provenance and Authenticity (C2PA), the open technical standard for embedding tamper-evident metadata in audio, image, and video files. Provenance is the foundation we have chosen for our Article 50 compliance. Relevant videos generated on the platform will carry C2PA-compliant signals that identify it as AI-generated and trace the chain of custody back to its source, so downstream platforms, factcheckers, and end users can verify what they are looking at using an open standard rather than a proprietary one.

We were also among the original group of companies that helped draft Partnership on AI's Responsible Practices for Synthetic Media, and we published a case study of our own implementation as part of that work. 

Both initiatives shaped how we think about disclosure, consent, and provenance, and these are the same building blocks the Code of Practice now formalizes at the European level.

The final Code is the product of three rounds of drafting and represents a real improvement on what came before. It is the result of months of consultation with industry, academia, and civil society. We participated actively throughout all three rounds of drafting, gave oral evidence and demonstrated our product in depth to the chairs and vice-chairs of both working groups, and submitted detailed written feedback at each round of revision. Our aim throughout was to keep the Code proportionate and technically workable, so that the transparency outcomes the AI Act wants to achieve happen in practice rather than collapsing under obligations no company, large or small, can actually implement.

We welcome several of the changes in the final version. The Code now acknowledges that videos produced by businesses for informational purposes circulate very differently from content on consumer-facing social platforms, and the obligations have been calibrated to reflect that distinction. 

Where startups are concerned, requirements such as providing a detection mechanism for AI content markings have been made proportionate to a company's stage and resources, which keeps the European AI ecosystem competitive without weakening the underlying protections. 

The final text is also clear that responsibility for the effectiveness, robustness, reliability, and interoperability of marking and labelling measures is shared across the entire AI value chain, from model developers through to deployers and distribution platforms. That is the only model that works in practice, because no single actor controls every link in the chain.

We are committed to continuing this work with the European AI Office in the months ahead, including on the guidelines for Article 50 that the Commission is due to publish in July. 

The Code and the guidelines together form the practical foundation for a transparent AI video ecosystem in Europe, and Synthesia is committed to demonstrating what that looks like in practice, proving that the right obligations, implemented well, make AI video more trustworthy, not less useful.

‍