I don't understand this point. If Google gave the data to OpenAI (which they surely haven't, right?), even then they'd not have consent from users.
As far as I understand it, it's not a given that there is no copyright infringement here. I don't think even criminal copyright infringement is off the table here, because it's clear it's for profit, it's clear it's wilful under 17 U.S.C. 506(a).
And once you consider the difficult potential position here -- that the liabilities from Sora might be worse than the liabilities from ChatGPT -- there's all sorts of potential for bad behaviour at a corporate level, from misrepresentations regarding business commitments to misrepresentations on a legal level.
They will gain a lot of lawsuit if they admit they trained on youtube dataset because not everyone gave consent.
But a lawsuit fails if essential elements are not met. If consent isn’t required for the lawsuit to proceed, then it doesn’t matter whether or whether not consent was granted. QED.
