OpenAI Supports EU Code of Practice on AI-Generated Content Transparency

The article exposes the fundamental tension between the rapid adoption of AI content generation tools and the ability of users to verify the authenticity of what they see online. Provenance signals like C2PA metadata are a promising foundation, but they are easily lost through resizing, screenshots, or platform re-encoding, making them insufficient alone. The ecosystem needs resilient, multi-layered approaches to prevent disinformation and protect election integrity, yet current techniques remain nascent and brittle.

OpenAI‘s support for the European Commission’s Code of Practice on Transparency of AI-Generated Content builds on years of internal work and ecosystem collaboration. Since 2024, the company has added C2PA metadata to DALL-E 3, expanded to ChatGPT and API outputs, and layered SynthID watermarks to provide complementary signals that survive more transformations. It also released a public verification tool (openai.com/verify), joined the C2PA Steering Committee, and contributed to drafting the Code alongside hundreds of stakeholders. This follows OpenAI‘s 2025 signing of the EU’s General-Purpose AI Code of Practice, showing a pattern of proactive engagement with European governance.

The practical takeaway for builders is that provenance must be treated as an ecosystem-wide infrastructure, not a single feature. No current method is foolproof, and reliability requires combining metadata, watermarks, product safeguards, and policy enforcement. As the Code is implemented, it should stay grounded in what works practically and remain flexible to technical limitations. Serious AI engineers should invest in interoperable standards like C2PA, contribute to open frameworks, and avoid assuming any one signal solves the trust problem.