Or it used something called an "analogy" which is a valid way to solve new problems.
replyThe very first line of the bug report:
reply> This is a bug found with our fuzzer: https://github.com/daedalus/fuzzer/
I think they're talking about the misconception that LLMs can only ever regurgitate their training data verbatim enough to constitute mass copyright violation. And that that's therefore "stealing"
reply