After the Mythos shock, how should security change?

Anthropic’s top model Mythos shocked the security industry by finding and analyzing real software vulnerabilities at very high speed.

A public version, Fable, was separated from the restricted partner-only Mythos, and U.S. export controls led to global service pauses and relaunches.

As AI can reproduce and analyze vulnerabilities far faster than humans, traditional patch SLAs and response playbooks are losing effectiveness.

Security teams now need to assess AI reproducibility, and access to frontier models has become a matter of national security and tech sovereignty.