> > Not really. The policy could take the view that the person > > instructing the LLM is the copyright holder because they had enough > > human interaction with the LLM to be able to claim copyright on that basis. > > That would provide a person to do (a) or (b). > > Are you sure?
Sorry, I was trying to be careful in my wording. I meant that the policy "could" take that view, not necessarily arguing that it *should*. I agree that whether it's correct or not is still unsettled law, though I believe it's correct. > I remember a case where, despite regular IP training, a colleague copied > a piece of GPL code wholesale into commercial software (as part of a > larger change). Fortunately that case was within the company's ability > to rectify. For copyrightable changes, how could someone using an LLM > be sure that that the LLM hadn't done the same? They couldn't, but I don't think that's the issue the policy is addressing. I think it's addressing the risk that the LLM output is uncopyrightable, not that it violates some other copyrighted file. As you pointed out, *that* risk is not unique to LLMs. A person could also inadvertently be infringing on a copyright or patent. > I suppose the policy also has to be future proof. The trend is clearly > to try to make LLMs do more with less coaxing, so the level of human > input is likely to decrease over time. Sure. But I think that, as long as it's modifying a piece of code, rather than writing its own, the work will still be a derived work of the original. But, as I said, there's no case law on the topic. If the ATT/UCB case had gone to trial, we'd have some, but it settled.
