Forking from a software with pirated elements is still retains its piracy unfortunately. Stack overflow comments are CC BY-SA 4.0 licensed, specifically to avoid the issue of piracy. I doubt everyone copying from it is properly attributing to the source, but thats much different then pretending the license doesn’t exist at all for personal benefit. AI is trained without restriction on available internet content, including github where many AGPL projects exist. Disclosure allows for a copyright respecting kernel to exist, still allowing pirates to fork it and add their stolen code if they want. Unfortunately the entire kernel is just setting itself up for future legal issues, like what originally handicapped BSD because AT&Ts copyright claim over Unix.