Is this a recall from memory or perfect understanding of context?

I haven't seen major debates in India on how ethical or legal it is to use "proprietary" data for training AI models. So I thought about bringing a LinkedIn discussion here. My stance is somewhat liberal with respect to data being used for AI training. I mean if data is available for public consumption, shouldn't it be considered passable for AI training? To look good in front of public, AI companies are making up things like "generative models don't memorise or reproduce training data" when in fact it depends on the prompt. I am against AI companies making up nonsense like this. But generally in support of publicly accessible data used for training purposes as the benefits outweigh demerits. What are your thoughts about this in general?