Yellow posted...
Yeah generative AI could destroy the entire internet.
It needs metadata flagging it as an ai image/video/sound. It then needs to be illegal to remove that metadata. You can even prove that it's AI-generated by including the hash of the model and settings used to create it so that it can be re-created.
Then browsers should tell you if it's AI-generated if you hover over it. This is the best solution I can think of.
Wouldn't work. One, you'd need new media formats with DRM to lock them down to avoid tampering with them. Then you'd need to have the issue of the formats needing to be supported by software, with the only "advantage" of them being DRM that limits their usability. Legality would be irrelevant unless you could get a worldwide implementation of the laws. And then there's the simple fact that you could to varying degrees just work around the DRM. For images in particular just open the image, tell your OS to take a screenshot, and save your new image having completely worked around the DRM.
There's also the issue that most people wouldn't bother checking a browser notification in the first place, and even if you got people trained to look, once people started bypassing the DRM and posting content without the "required" metadata people would get the false impression that the content is genuine, even if it was reasonably obvious it was AI generated. Which basically makes the problem worse than what it would be without the metadata in the first place.