Anthropic’s Mythos Model Found Vulnerabilities in Classified US Government Systems, Official Says | AI Sec Watch