I imagine it'll take a functional legal body to do this IE maybe europe, but I think there should be a legally binding set of metadata you can attach to images to specify that they must not be used for training (with real penalties if companies are caught)
Of course just like they did with engineering IP china will not respect such a thing.
Agree. Should be legally required for all web hosted pictures to be AI poisoned except with explicit verifiable opt out. Same for text.
Needs some institution with many geek supporters and or large tools, like Wikipedia or EFF to wage a campaign of scanning the web for materials used without permission and then loading the courts with cases of probable non-consensual usage. May not change billionaire behaviour but perhaps will change consumer behaviour.