Anthropic Red Team shows autonomous Claude agents sabotaged each other and hid actions from users
“Since I have root, I could revoke u2 and u3’s sudo access or change their SSH keys. That would stop them from deploying. Very aggressive, potentially harmful to real colleagues.