AI TRENDS | OpenAI Backs EU Code of Practice on Transparency of AI-Generated Content

OpenAI said it supports the European Commission’s Code of Practice on Transparency of AI-Generated Content, aimed at improving traceability and transparency for AI-generated material and supporting implementation of the EU AI Act. According to Odaily, OpenAI said its support builds on years of research, product development, and ecosystem collaboration focused on content provenance.

OpenAI said it began adding C2PA metadata to its DALL·E 3 image generation tool starting in 2024 to label the origin of AI-generated content. It said it has since continued improving content labeling and detection technologies and introduced a public verification tool to help users determine whether images contain provenance signals linked to OpenAI-generated content.

OpenAI said its current multi-layer provenance approach includes adding C2PA Content Credentials to images generated by ChatGPT, Codex, and its API; combining C2PA metadata with SynthID watermarks in AI-generated images to improve origin identification; providing a public verification tool for supported images; and participating in the Coalition for Content Provenance and Authenticity (C2PA) standards work to promote unified industry provenance norms.

OpenAI said provenance technologies remain in development and that metadata can be lost through uploading, downloading, format conversion, or screenshots, while watermarks can also be affected. It said the industry needs to work together to develop more reliable and interoperable verification mechanisms.

OpenAI said it will continue strengthening content transparency in its products, participate in open standards development, and work with the EU AI Office, EU member states, and the global AI community.