I also had to argue with a lawyer on the same point - he held a firm belief that “Modern GenAI systems” are different from older ML systems in that they are non

I also had to argue with a lawyer on the same point - he held a firm belief that “Modern GenAI systems” are different from older ML systems in that they are non-deterministic and random. And that this inherent randomness is what makes them both unexplainable (you can’t guarantee what it would type) and useful (they can be creative).