Anthropic Mythos just broke the four-minute mile in cyber offense

This briefing argues that recent AI models (e.g., Anthropic Mythos) have accelerated autonomous discovery and exploitation of zero-day vulnerabilities—producing numerous working exploits and collapsing time-to-exploitation—urging CISOs to adopt immediate measures: integrate AI security agents into code and CI/CD, form governance teams, enable continuous patching, build automated response capabilities, and establish measurement frameworks to ensure resilience before these offensive capabilities are widely weaponized.