Hacker News Reader

That doesn't mean that all of the output from an LLM trained on GPL code is a derivative work (and therefore GPL'd too).

A model that provably engages in systematic, difficult-to-detect plagiarism must itself be considered plagiaristic.