On Fri, Jul 24, 2026 at 10:41:10PM +0300, Aigars Mahinovs wrote: > AFAIK Bartz v. Anthropic is one such landmark case - [2] > https://legalblogs.wolterskluwer.com/copyright-blog/the-bartz-v-anthropic-settlement-understanding-americas-largest-copyright-settlement/ > has some analysis and background, for example. > > The key finding, for our case was the ruling from last year: "that > Anthropic's use of legally acquired books for AI training was > "quintessentially transformative" and protected as fair use". > > This basically is a bedrock decision that established that copyright of the > training materials do NOT survive the transformation inherent > in the learning process and thus the output of an LLM is NOT a derivative > work of the training material in the legal sense.
While very important in many ways, that court decision is strictly about the impact of copyright on the *training* part of modern LLM-gen-AI systems, and does establish what you quoted in the second paragraph above. But, AFAIU, it does not establish anything about the copyrightability of / potential copyright infringement in the *output* of LLMs --- I was already aware of this, but the summary you quote above says so too: "Limited scope: Crucially, this settlement […] does not: […] Cover claims based on AI outputs that might infringe copyrights." Cheers -- Stefano Zacchiroli - https://upsilon.cc/zack Full professor of Computer Science, Polytechnic Institute of Paris Co-founder & Chief scientific officer Software Heritage
