OpenAI has incorporated specifications from the Content Authentication Alliance (C2PA) into DALL-E 3. All images generated by the DALL-E 3 serving API or ChatGPT will include a visible watermark to identify the image as AI-generated.

The watermark will include details like the date the image was created and the C2PA logo in the upper left corner. This is intended to make it clear to users whether the image was created by a human or AI. OpenAI says the watermark won't affect the quality or speed of image creation, but can increase file size by 3-5% via the API and 32% when created using ChatGPT.

However, there are still ways for users to remove the AI ​​provenance of images. According to OpenAI, cropping or filtering the DALL-E output can remove this provenance data.

Metadata like C2PA can easily be removed, either accidentally or intentionally. For example, most social media platforms today remove metadata from uploaded images, and actions like taking a screenshot can also remove that metadata.

Microsoft is also adopting the C2PA specification for Bing Image Creator output, the company said: "AI-generated images generated by Bing Image Creator now include an invisible, digital watermark that complies with the C2PA specification."

Meanwhile, Meta just announced that it will begin labeling content uploaded to Facebook, Instagram, and Threads if it was generated using AI. The move is part of an ongoing effort to develop industry standards for transparent labeling of AI content.