But this doesn’t solve the problem because a nefarious actor would simply use a service that doesn’t watermark their images. That is to say, if an image has a watermark it is definitely AI. If it doesn’t have a watermark, you can’t determine whether or not it’s AI.
Wish granted: https://deepmind.google/models/synthid/
There are several others.
But this doesn’t solve the problem because a nefarious actor would simply use a service that doesn’t watermark their images. That is to say, if an image has a watermark it is definitely AI. If it doesn’t have a watermark, you can’t determine whether or not it’s AI.