Technical analysis suggests that AI-generated images contain hidden steganography in the yellow channel, encoding prompts and user data. This metadata may be used by AI companies to identify synthetic content and prevent "model collapse" caused by training new models on AI-generated data.
