Comment by finolex1

The issue is what happens if/when the models grow capable enough that the providers can't stop them even if they want to. You could have strict penalties but that's not going to solve an open research question.