The really weird thing about Facebook's AI training crawler ignoring robots.txt is their link preview process (with a different user-agent) respects the robots.txt so hard that even when I exempt it, it still thinks it's forbidden. 🙄