It's not about "banning all AI," but trying to distinguish between using Al assistants in everyday development and feeding source code into someone else's model training systems.