OpenAI Tags AI Art to Combat Misinformation
OpenAI will add watermarks from the C2PA, including invisible metadata and a visible CR symbol, to images generated by ChatGPT and DALL-E 3 to help identify AI-generated content.
Summary
- OpenAI will add C2PA watermarks to DALL-E 3 images starting Feb 12 on mobile and on the ChatGPT site. The watermarks help identify AI-generated content.
- The watermarks include invisible metadata and a visible CR symbol in the top left corner showing the content is AI-generated.
- The metadata can be checked on sites like Content Credentials Verify to see which AI tool created the content.
- The C2PA watermarking initiative, supported by companies like Microsoft and Adobe, aims to tag AI-generated content to address misinformation concerns.
- However, OpenAI notes the metadata can still be easily removed, so it's not a perfect solution. Posting images on social media often strips metadata.
- OpenAI believes adopting signaling methods like watermarks, despite limitations, helps increase trust in digital information.