A number of gen AI companies have suggested they’re surprised they’re being called out for training on copyrighted work without consent, as if this would be a rewriting of copyright law and an unfair change of course they couldn’t have predicted.
This is not true. First, it’s not a change in the law, but an application of existing law. Second, they have always known they would be called out on it one day. I know - I remember looking at this exact question eight years ago when I was running one of these companies.
They have always known there was a risk their training in this way would be shut down. Regulators shouldn’t be convinced into thinking otherwise.