I think it's a reference to Google's book scanning product, which is structurally similar: they use copyrighted material to provide a new kind of service, which contains an echo of the original material. The book scanning and the related search product is supposedly legal under U.S. copyright law.
The point is that GP has no reason to believe Google, like Meta, also used copyrighted materials for training its AI. Why did Sarah Silverman sue OpenAI and Meta but not Google?
I didn't accuse Google of using copyrighted materials for training its AI. I accused Google of existing under a different set of laws than mere citizens.
As an example, the mass usage of copyrighted materials to build Youtube.
It didn't.
YouTube respects creators' copyright in the videos and pays them.