OpenAI Signs EU Code of Practice on AI Content Transparency

OpenAI has signed the EU Code of Practice on AI content transparency, a voluntary framework developed under the EU AI Act. The Code focuses on provenance standards, meaning technical methods that help identify whether content was made by an AI system. By signing, OpenAI commits to implementing and supporting tools that label or watermark AI-generated text, images, audio, and video, so that people encountering that content can understand its origin. The Code is part of a broader EU effort to establish common industry standards before binding rules take full effect.

Journalists and fact-checkers who need to verify whether images, video, or text submitted to them were AI-generated. Platform trust and safety teams responsible for detecting synthetic content at scale. Policy and compliance officers at media organisations operating under EU rules. Researchers studying the spread of AI-generated material online.

Signing the Code gives OpenAI a formal seat in shaping how provenance standards develop across the EU market, rather than having those standards handed down to it later. For organisations that publish or moderate content, it signals that major AI providers are moving toward interoperable labelling systems, which could reduce the patchwork of incompatible detection approaches currently in use. If the Code's standards are adopted consistently, tools built to check content provenance would need fewer workarounds to function across different AI platforms.

This is a voluntary commitment, not a binding legal obligation, and the Code sets no enforcement mechanism for signatories who fall short. Provenance tools, including watermarking and metadata tagging, are currently easy to strip or alter, so their practical value for detecting manipulated or re-exported content remains limited. Signing also does not mean OpenAI's existing products already meet the Code's standards. The gap between agreeing to a framework and shipping working implementations at scale can be wide and slow to close.