Anthropic withholds powerful Claude Mythos model from public release due to dangerous cybersecurity capabilities, instead sharing it with select security partners to patch vulnerabilities.

Anthropic's new Claude Mythos model possesses advanced cyber capabilities that are deemed too dangerous for public release until critical software infrastructure is strengthened

The company plans limited release to key cybersecurity partners to identify and patch vulnerabilities in major software systems at scale

Anthropic is attempting to close a critical security window by giving defenders access to the technology before entering a new era of AI capabilities

The decision reflects a strategic approach to responsible AI deployment, prioritizing infrastructure security over broad public access