Post #2900976

2026-04-19 21:44 UTC

@bkuhn@fedi.copyleft.org @cwebber@social.coop @ossguy@fedi.copyleft.org @richardfontana@mastodon.social @evan@cosocial.ca There's multiple undecided cases in Europe which show that the possibility that training violations of copyright may be very much a thing in the EU. As an example: https://www.ipcuria.eu/case?reference=C-250/25 Due to this, the EU government is debating some potentially very strong regulations on model training, including a requirement to pay all copyright holders before ingestion: https://www.osborneclarke.com/insights/eu-copyright-law-and-generative-ai-watershed-moment ...

Replies (2)

@fuzzychef@m6n.io 2026-04-19 21:46

@bkuhn@fedi.copyleft.org @cwebber@social.coop @ossguy@fedi.copyleft.org @richardfontana@mastodon.social @evan@cosocial.ca ... no existing commercial AI tool complies with these prospective regulations and potential court precendents, which means that should things go the way of strong regulation and limits on model training -- or in the direction of transfer of copyright obligations to the user -- we'll be ripping out a bunch of stuff from OSS projects.

Open ##2900977
@bkuhn@fedi.copyleft.org 2026-04-19 23:28

@fuzzychef@m6n.io This doesn't contradict anything I said, and I'm aware of the cases. No case anywhere has been decided yet, and compulsory licensing (which is anti-copyleft) is a dangerous potential outcome that regulators and judges around the world tend to favor. @cwebber@social.coop @ossguy@fedi.copyleft.org @richardfontana@mastodon.social @evan@cosocial.ca

Open ##2900978