Mythos Preview Builds PoC Exploits in Automated Vulnerability Research

Cloudflare tested Anthropic’s Mythos Preview, a security-focused AI, and found it can automatically find vulnerabilities, construct exploit chains from low-severity primitives, generate and iterate PoC code, and substantially reduce triage time; the report warns this capability is dual-use, increasing attacker effectiveness and emphasizing the need for architecture-level defenses and consistent safeguards.