Anthropic on Tuesday announced an expanded three-tier Cyber Verification Program that opens its most powerful Claude AI models — Opus 5.5, Sonnet 5.5 and Mythos 5.1 — to vetted cybersecurity professionals with reduced safeguards.
The program’s tiers are Defense, Red Team, and Specialized, with the latter reviewed alongside the U.S. government for critical-infrastructure testing. The folded-in Project Glasswing surfaced at least 129,000 verified software vulnerabilities between April and July 2026, plus another 5,500 through open-source scanning — with more than 33,000 rated critical or high severity.
The move positions frontier AI models as tools for defenders as well as researchers stress-testing software, at a moment when AI-assisted vulnerability discovery is accelerating on both sides.
The announcement was first reported by Reuters on Oct. 6 and covered widely across the security press on Wednesday.
