Overview
- Anthropic announced a merged Cyber Verification Program that folds Project Glasswing into a single three-tier system and is now accepting applications for Defense Access, Red Team Access and Specialized Access.
- All three tiers grant access to Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1 with different eligibility rules and technical controls for defenders, penetration testers and organizations working on safety-critical systems.
- Specialized Access will be limited and vetted in collaboration with the U.S. government while existing Glasswing members are being migrated into that tier without reapproval for current models.
- Program participants must accept data retention so Anthropic can monitor for misuse and the company plans Enterprise Frontier Safeguards later this fall to let eligible customers store data in cloud infrastructure they control, with Bedrock access restricted to those eligible.
- Anthropic also released Claude Haiku 5.5 as a lower-cost, faster model, and the company points to large-scale vulnerability findings from its partners—about 129,000 verified issues plus roughly 5,500 from Anthropic with more than 33,000 rated critical or high—to argue defenders need stronger AI tools as openness from other vendors increases competitive pressure.