> we should've caught this behavior earlier

Who's "we"? Real Anthropic employees or the bots you allow to run free 24/7?

What countermeasures have you/your bots come up with to avoid disclosing user information going forward?