It’s still debatable if training a computer neutral network on public data is 'wrong' when we very much accept it as a right for biological neural networks.
It's even less worthy of sympathy - like a counterfeit piece of art being counterfeited. And there isn't even an original, just like a made up counterfeit.