Post #3043073

2026-05-18 08:30 UTC

@muvlon@hachyderm.io Interesting! I did indeed only see media reporting on it, but did not read the actual case. Will read up on (I guess) [this document](https://media.cadc.uscourts.gov/opinions/docs/2025/03/23-5233.pdf) to get a better understanding. Though I hold that EU should not be subject to US rulings, it is helpful to see how other jurisdictions interpret it as a hint of how EU might handle it. Either way, my personal belief is still that LLM output is subject to the copyright of its training material, and thus 99% unfree.

Replies (1)

@muvlon@hachyderm.io 2026-05-18 10:58

@allpurposemat@mastodon.gamedev.place Oh yes, EU is different still and I don't know of any ruling in any particular direction there yet. All I wanted to say was that even in the US, it's not clear yet. I think it should also be noted that in the Thaler case, there was no LLM or other modern "genAI" involved. He initially submitted his request in 2019, years before GPT-3 was released and started the current AI craze. Thaler just developed a generative algorithm that outputs art, AFAICT it doesn't even use a transformer architecture or massive amounts of stolen training data. In my layperson opinion, this is pretty squarely classic "computer aided art", not slop, and thus copyrightable. He just messed up by trying to attribute it to the algorithm instead of himself.

Open ##3043074