Anthropic is expanding its Cyber Verification Program (CVP), which allows vetted cybersecurity professionals to test its most capable AI models with reduced safety guardrails. The initiative integrates the company’s previous efforts, including Project Glasswing, which focused on securing critical global software.
Between April and July, partners in Project Glasswing identified at least 129,000 verified software vulnerabilities. Additionally, Anthropic’s own open-source scanning efforts uncovered 5,500 more vulnerabilities between April and October. Over 33,000 of these findings were classified as high-severity or critical.
The new CVP framework consists of three tiers, each with specific vetting requirements and safety controls. Participants can access Claude Opus 5.5, Claude Sonnet 5.5, and Claude Mythos 5.1 models. The tiers range from incident response and malware analysis to authorized penetration testing and red teaming.
The most restricted tier is reserved for a select group of organizations authorized to test critical systems, such as power grids, aviation flight systems, and interbank transfer infrastructure. Anthropic reviews all applicants in collaboration with the United States government.
