OpenAI announced support for the European Commission Code of Practice on Transparency of AI-Generated Content and tied it to C2PA provenance work.
What OpenAI Signed On To
OpenAI has announced its support for the European Commission's Code of Practice on Transparency of AI-Generated Content. A code of practice is a voluntary framework: it sets expectations for how organizations disclose when content has been produced or altered by AI, without carrying the force of a statute on its own. By backing it, OpenAI signals that it intends to align its labeling and disclosure practices with a shared industry baseline rather than a proprietary one.
The practical effect is about interoperability. When multiple model providers, platforms, and publishers commit to the same transparency expectations, a label applied by one system means the same thing when it reaches another. That consistency is what makes disclosure useful to a reader, who otherwise has no way to tell a hand-drawn image from a generated one, or an edited recording from an original.
The C2PA Provenance Connection
OpenAI tied this support to its work on C2PA provenance. C2PA is an open technical standard for attaching provenance information to a piece of media — a tamper-evident record of where the content came from and how it was modified along the way. Instead of relying on a visible watermark that can be cropped out, provenance data travels with the file as signed metadata that downstream tools can read and verify.
Pairing a transparency code with a provenance standard matters because the two solve different halves of the same problem. The code establishes *what* should be disclosed; C2PA establishes a concrete, verifiable *mechanism* for carrying that disclosure. A commitment to transparency is only as strong as the technical plumbing behind it, and provenance metadata is the part that survives copying, re-uploading, and format conversion when it is implemented carefully.
How Provenance Labeling Works in Practice
For a team building or distributing AI content, provenance-based transparency generally comes down to a few steps that fit into an existing publishing pipeline:
- Generate or edit the asset, then embed signed provenance metadata describing that it was AI-produced or AI-modified.
- Preserve that metadata through storage, editing, and export so it is not silently stripped by intermediate tools.
- Expose a way for end users or downstream platforms to inspect and verify the credentials, rather than trusting a claim at face value.
The hard part is usually preservation, not creation. Many content platforms and editing tools discard extra metadata by default, so provenance can be lost between the point of generation and the point where a reader actually sees the file. Broad adoption of a shared standard is what pressures those intermediate tools to carry the data through instead of dropping it.
Why the Alignment Is Worth Watching
Voluntary codes and open standards tend to reinforce each other. A code gives providers a common target to aim at, and a standard like C2PA gives them a shared way to hit it, which lowers the cost for the next organization to participate. When a major model provider commits publicly, it raises the baseline expectation for peers and reduces the space for opaque, unlabeled generation.
The open questions are about coverage and durability: whether disclosure survives the full path from creation to consumption, how verification is surfaced to ordinary readers, and how consistently participants apply the practices they have endorsed. Transparency is a process to maintain, not a checkbox — and provenance is the part that makes it checkable.