[yuuuuyu]

If you are holding copyright to something, it will be on you to prove it's in there.

[manojlds]

I am obviously clueless, but if it's a case, can't one demand what the training data is?

[arthurcolle]

Based on what case law? In what jurisdiction?

[nwienert]

It's very easy to show the models generating near-identical data to copyrighted data, which is enough to get courts to force them to allow discovery.

[yuuuuyu]

This not have happened all over the place yet is evidence against this being as easy as you make it sound.