Yeah it is, thats why you’ll hear ‘cleanroom development’ tossed around, to refer to someone that has never seen the other software’s source is developing its replacement. This matters legally because its an airtight way to prove no copyright could have been violated. Usually the community overlooks such minor violations to support new developers, but unfortunately this sentiment is changing fast thanks to the modern piracy landscape. The term ‘cleanroom’ has gotten pretty muddied with AI taking it over to mean ‘mathematically exclusively derived from the original source, but cleanroom makes us claim legal overship over someone else’s work so that’s what it is’. You can see how the meaning has drifted quite a lot once it reached marketing. I haven’t seen any precedent in US or other courts on the originality of AI code, but if you have any links I’d be interested in learning more. It would be quite a monumental ruling, because it requires us to ask - how small of an AI model is technically original? If I train an ‘AI’ on a single pice of work, can I then claim ownership over it legally? Or does it specifically require whatever the current state of the art datacenters are at? Cause then its just legal for cooperation’s to ignore copyright and no one else.