Well I hope the community decides not to press the issue for a good while, lulling companies into a false sense of security that they can release large high-quality models with "licenses" attached, before the seminal court case that rules that language models are not in fact copyrightable and no one can stop us doing anything we like with such data.