Project maintainers know what level of AI use is acceptable for themselves, and quantifying it and enforcing it for random contributors is very difficult.