As part of its push to implementing measures to ensure that users can differentiate AI-created content from those done by actual humans, OpenAI has updated its DALL-E image generator to offer more transparency about computer-generated images.
OpenAI has incorporated specifications from the Coalition for Content Provenance and Authenticity (C2PA) into DALL-E 3. Now, images produced by the API serving DALL-E 3 or ChatGPT will have a visible watermark, and embedded metadata to indicate that they are AI-generated. This change is also set to rollout to mobile users by February 12.
The visible watermark will show details including the date the image was created and the C2PA logo in the upper left corner. This is intended to clearly indicate to users whether an image was created by a human or an AI system.
Images generated via the API will contain a signature to indicate that they were created by the DALL-E 3 model. On the other hand, images generated from ChatGPT will have an additional manifest to indicate that they were created via ChatGPT.
OpenAI states that the watermark will not reduce image quality or creation speed but may increase file sizes by 3-5 percent via API and 32 percent when generated in ChatGPT.
While definitely a step forward, it’s not a perfect solution to address concerns regarding provenance. Most social media platforms remove metadata from uploaded images, and taking a screenshot can also result to an image with its metadata removed, leaving you guessing whether it was AI-generated or not.
Emman has been writing technical and feature articles since 2010. Prior to this, he became one of the instructors at Asia Pacific College in 2008, and eventually landed a job as Business Analyst and Technical Writer at Integrated Open Source Solutions for almost 3 years.
