Not necessarily. If it turns out, for example, that language generation models trained on GPL code must be GPL, then it means that there's a possible path to more open models, if content creators continue creating copyleft content ecosystems.