The article is right. If no human effort is involved then the code should be open source. The article also implies a very detailed line too. It doesn’t say that the human effort part can’t be copyrighted or intellectual property.
As an example, I have developed a set of grounding guidance for my AI models at my work. The specific grammar and vocabulary belong to my employer. But I also have a clean-room version of it that completely strips out that proprietary infringing information while still retaining a similar semantic structure. That will be posted next month on my GitHub under a GPL-3 license like I do with my other non-infringing AI-assisted work.
As an example, I have developed a set of grounding guidance for my AI models at my work. The specific grammar and vocabulary belong to my employer. But I also have a clean-room version of it that completely strips out that proprietary infringing information while still retaining a similar semantic structure. That will be posted next month on my GitHub under a GPL-3 license like I do with my other non-infringing AI-assisted work.