1 article on this topic.
OpenAI's new image models generate faster, more realistic output and pair it with C2PA metadata plus a new SynthID watermark — but the company's own safety data shows the misuse rate isn't zero.