@Ganneff Personally I watched it from the sidelines, and while I am politically very much against AI usage in the project, here it could be rebuffed with the copyright:

There's no way that LLM generated code can be copyrightable when there is no free LLM with its training data available
I read which model they used, and that certainly doesn't fulfill the DFSG. If we do not care about that then we might as well ship proprietary firmware in main.