@marcel @helma @molly0xfff Let’s be clear though: there is still a mountain of difference between, say, an AI company training their proprietary, commercial model on MIT-licensed code (perfectly legal) and doing so on AGPL-licensed code (breach of license). Let’s not abandon the tools we have. Just because they’re breaking the laws and contracts that exist today doesn’t mean we should abandon them for future ones that they may or may not be bothered to obey. You don’t let a thief define what does or doesn’t constitute theft. You enforce the law. That said, I realise how utterly naïve that sounds as we stand mired in technofascism. So maybe what we do is destroy the looms.

@aral @marcel @helma if AI training is fair use, then training on AGPL licensed code is permissible

@claudius @aral @marcel @helma commercial use doesn't prevent it from being fair use. see for example the Google Books case

@molly0xfff @aral @marcel @helma It's a big part in the decision, though, as far as I know. Who benefits matters a lot.

@claudius @aral @marcel @helma right, it is a factor