OpenAI has announced that it will add watermarks to the metadata of artificial intelligence (AI)-generated images created by DALL-E 3. The company said it will now use the open technical standard adopted by the alliance for content provenance and authenticity. C2PA), and add information that the image was created by AI, the name of the AI tool, and the name of the app used to create the image to its metadata. The move comes as Meta announced it needs to adopt a common standard to help AI companies identify and label AI content on their social media platforms.
In a post, OpenAI revealed the move as well as its technical details. It added that images generated with ChatGPT on web clients and APIs, which use the DALL-E 3 model, will now include a new metadata according to the C2PA standard. The same watermark process will be started on ChatGPT app by 12th February. The C2PA standard is a specific watermarking technology that adds a stamp to the image itself and also embeds the information inside the image. As a result, a CR symbol can be seen at the top left of the image and a detailed version can be checked in its metadata.
Through metadata, users can investigate the origin of the image, including information about the AI model and app used to create it. In the examples shared by OpenAI, the metadata shows a content summary that says, “This image was created with an AI tool.” A separate tab for Process shows whether the API, web client, or ChatGPT was used, in addition to showing the underlying AI model. According to the company, adding metadata may slightly increase the image size, but will not affect the quality.
Although this makes the image more secure than simply a visual marker, there are ways to bypass it. OpenAI highlighted that many social media platforms remove metadata from an uploaded image, and taking a screenshot of the image will also remove it. Therefore, this method may not be sufficient to determine whether an image was actually created by DALL-E 3 or other AI models.
C2PA includes companies like Adobe, Microsoft, BBC, Sony, Leica, Nikon and others. It is pushing for the adoption of this technology as a method to detect and accurately label AI-generated content. The CR mark created by Adobe was also given by this group.