The important part is the "neither intentional nor wanted". That's why you can treat potential license-violating output from an LLM similar to from a human. It might happen, but the likelyhood is low and you'll have to have systems in place for verification regardless if you're using LLMs or not.
(Oh and yes, many anti AI fanatics believe .that an LLM can only produce output that replicates what was in their training data exactly and nothing else)
mbpaz (npub1xcd…hnp8) Chad McCullough (npub1ded…kves) elementary (npub16nm…cc46)
