I think there is a very practical reason to attribute AI contributions: AI models are improving in ability. Being able to know when and what contributed the code, would allow people to more easily deploy newer AI to examine the work of previous AI, to improve or replace it. Plus, some AI will likely be specialized in specific domains, so you wouldn't want different agents from stepping on each other's toes. Something oriented around GUI design, probably shouldn't be handling graphics optimizations.

This removal of authorship will just make things more difficult in the long run.