OpenAI has strengthened its system for verifying the provenance of AI-generated images. The company announced its status as a C2PA Conforming Generator Product and unveiled an early version of a public verification tool.
OpenAI has also begun adding invisible Google SynthID watermarks to images generated in ChatGPT, Codex and through the API. The service looks for Content Credentials and SynthID to help determine whether an image is linked to OpenAI tools.
The company specifically notes that the absence of a signal does not definitively rule out such a link: metadata and watermarks can disappear when images are saved again, converted to another format or captured in screenshots.
Meanwhile, Google announced that OpenAI, Kakao and ElevenLabs are expanding their use of SynthID in their AI-generated content, with checks for these signals coming to Gemini, Search and Chrome.
