Unauthorized Group Claimed to Have Gained Access to Anthropic's Restricted Mythos Cyber AI

TechCrunch reports that an unauthorized group has reportedly gained access to Anthropic’s Mythos, the company’s restricted cybersecurity AI model. Anthropic confirmed it is investigating the claims but stated there is currently no evidence that its internal systems have been compromised.

Mythos is positioned as a highly capable offensive security tool — its use by unauthorized parties raises direct dual-use concerns, as a model purpose-built to discover and potentially exploit vulnerabilities could be wielded by actors outside Anthropic’s controlled access program. The incident is still developing; practitioners working with or alongside Mythos integrations should watch for further updates from Anthropic.