OpenAI has made it official that it will apply labels to AI-generated images created with its DALL-E 3 model. Specifically, the company will apply watermarks to image metadata as companies implement support for AI standards Coalition for Content Provenance and Authenticity (C2PA).
The C2PA watermarks will appear in the images generated on the ChatGPT site and on the API for the DALL-E 3 model. As regards mobile, however, They will arrive by February 12th. Getting even deeper into the news, the watermarks will consist of both invisible metadata and a visible CR symbol, which will appear in the top left. Users will thus be able to verify the origin and which AI tool was used to create the content. Additionally, watermarks will not be applied to videos and text content.
Finally, OpenAI highlighted that adding watermark metadata to images is a “negligible effect on latency and will not affect image generation quality.” By the way, it will also slightly increase the image size for some tasks.
