Anthropic expands verified cyber access across three Claude tiers

Three levels of verified access

Anthropic expanded its Cyber Verification Program on October 6, bringing it together with Project Glasswing in a three-tier system for vetted security work. The company says each tier can provide access to capabilities in Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1, with different verification requirements and cyber controls. Defense Access covers activities such as incident response, malware analysis and vulnerability validation; eligible applicants may include smaller security teams and individual researchers with a history of reported vulnerabilities. This is an application process, not automatic access for every researcher.

Red Team Access adds authorized penetration testing and is currently available only to organizations. Anthropic says it will still block actions that could cause physical harm or mass disruption. Specialized Access, with the fewest cyber blocks, is reserved for a limited group testing high-risk systems such as power grids or flight operations. Anthropic says it reviews those organizations with the US government. Its public models retain their usual cyber safeguards. The company reports positive results from internal evaluations and prior Glasswing work, but the announcement does not independently establish how the expanded program will perform or prevent misuse at scale.