The European Commission's Code of Practice for marking and labeling AI-generated content is expected to be finalized this month, June 2026, giving developers, deployers and platforms concrete guidance ahead of the EU AI Act's August 2, 2026 transparency deadline. The Code is a non-binding implementation document, but in practice it will set the bar regulators expect industry to meet.
The Code covers three connected questions. First, what counts as AI-generated or AI-manipulated content for the purposes of disclosure — including images, audio, video and synthetic text. Second, which technical methods (watermarking, provenance metadata, cryptographic signatures) are acceptable for labeling that content. Third, how those labels should be surfaced to end users in a way that is intelligible without being intrusive.
For frontier labs and major platforms, the Code formalizes a direction that many have already been investing in. Several large model providers have committed to provenance standards such as C2PA, and platforms have begun testing visible and invisible markers on AI-generated media. The Code is expected to align expectations across the industry rather than introduce surprises.
The August deadline matters because it is when the EU AI Act's transparency obligations begin to apply more broadly to general-purpose AI systems and to high-risk systems already in scope. Companies operating in the EU need to be able to demonstrate not only that they label AI-generated content where required, but also that their labeling regime is technically robust and survives common transformations.
The Code arrives in a global context of growing convergence on content provenance. The U.S. has been moving in fits and starts on labeling standards, several U.S. states have passed disclosure laws aimed at political deepfakes, and platforms operating worldwide are increasingly unifying their labeling and provenance pipelines to satisfy multiple jurisdictions at once.
For AI builders, the practical implication is that watermarking and provenance infrastructure should be treated as a first-class product surface, not a compliance afterthought. The EU's Code is the most explicit signal yet that this is where the regulatory consensus is heading.
Source: [DLA Piper](https://www.dlapiper.com/en-us/insights/publications/2026/03/white-house-releases-the-national-policy-framework-for-ai-key-points)