If their AI model is good enough to make PRs, it's good enough for them to ask it about standards and norms. They can't be both massively enabled and helplessly unaware.