TRPL: A copylet license that treats API usage as derivate works

> If you train on TRPL code, the model weights and training data become "Functional Output" and must be shared. No black box models.

This doesn't work because AI companies don't accept licenses. Either the training is already infringing (in which case this doesn't matter much) or it's fair use/permitted by law (in which case this also doesn't matter much).