OpenAI’s DALL-E 3 image generator will automatically add watermarks to image metadata to comply with Content Provenance and Authenticity (C2PA) Coalition standards.
Metadata will be used to tag images created by artificial intelligence on the ChatGPT website and when connecting to the API for the DALL-E 3 model, OpenAI said. When displaying such images on resources that support Content Credentials technology, an image with the symbols “CR” will be displayed in the upper left corner of each image. This will allow you to establish the origin of the file, but for now only pictures, not videos or text, will be marked. Adding metadata will only slightly affect the speed of the system and will not affect the quality of the image in any way, OpenAI emphasized.
However, watermarks will not be a reliable way to combat disinformation. OpenAI notes that C2PA metadata is “easy to remove, either accidentally or intentionally,” especially since most social media platforms often strip metadata from uploaded content. They are also not read when taking a screenshot.
