I think the real benefit for this is that probably that it establishes trust as the default, and acts as a discriminator for good-faith uses of "AI". If most non-malicious uses of ML are transparently disclosed, and that's normalized, then it should be easier to identify and focus on bad-faith uses.