By that logic, LLMs that were trained purely on proprietary code should be allowed to generate proprietary code (Microsoft, Oracle, Apple, etc. should have a quite large codebase when combining all their products).
And then again, I wonder which license should be used. Is this a case of copyleft, and therefore everything generated should be something like GPL, or is it more like public domain because no copyright can be exercised?
And then again, I wonder which license should be used. Is this a case of copyleft, and therefore everything generated should be something like GPL, or is it more like public domain because no copyright can be exercised?