Are you saying that Claude will help you perform malicious attack against infrastructure if you ask it to and that anthropic should be able to stop that? I could see reasonable use cases for this like penetration testing against your own infrastructure. That’s not the same as making weapons or meth.