Particle.news

Anthropic Recasts Cyber Verification Program Into Three Tiers

Vetted security teams will gain controlled access to Claude Opus 5.5, Sonnet 5.5, Mythos 5.1 under U.S. government vetting with layered safeguards.

Anthropic's logo, a keyboard and a robotic hand are shown in this illustration created on June 5, 2026. REUTERS/Dado Ruvic/Illustration/File Photo

Overview

  • Anthropic announced the revamped Cyber Verification Program on Tuesday, Oct. 6, 2026, merging Project Glasswing and its earlier CVP into Defense, Red Team, and Specialized tiers.
  • The program gives qualifying organizations access to Claude Opus 5.5, Sonnet 5.5, and Mythos 5.1 so they can perform incident response, malware analysis, and authorized penetration testing.
  • Anthropic will relax some high‑risk dual‑use classifiers for verified CVP users and will keep universally prohibited activities blocked for all users.
  • Project Glasswing partners reported at least 129,000 verified vulnerabilities found between April and July and Anthropic’s own scans added about 5,500 more through October, with more than 33,000 rated high or critical.
  • Internal CyScenarioBench testing shows Defense Access blocks most complex offensive scenarios while Red Team Access completed 34 of 50 assessed tasks, highlighting trade‑offs between safety controls and offensive capability.