Anthropic said on October 6 it is expanding its Cyber Verification Program, allowing vetted cybersecurity professionals to test its most powerful AI models with fewer safeguards. The company said its Project Glasswing initiative helped uncover more than 100,000 software vulnerabilities this year. Glasswing partners found at least 129,000 verified vulnerabilities between April and July, while Anthropic’s open-source scanning found 5,500 more between April and October. More than 33,000 vulnerabilities were rated critical or high severity. Anthropic said the figures likely undercount the total and estimated the true impact could be at least five times higher, because the data came from a survey of a limited number of partners.

The revamped program combines two programs Anthropic had operated for six months. Glasswing gave organizations protecting critical software access to Claude Mythos, Anthropic’s most cyber-capable model family; the earlier Cyber Verification Program gave vetted security teams reduced safeguards on Claude Opus and Sonnet. Anthropic’s April announcement of Claude Mythos Preview had raised fears that AI could hack software before it was secured.

The expanded program has three tiers, each with its own verification requirements and security controls, and all include Claude Opus 5.5, Sonnet 5.5, Mythos 5.1 and future models. The Defense tier covers incident response and malware analysis, and accepts applications from security teams, critical infrastructure operators, open-source maintainers and researchers with a record of reported vulnerabilities. The Red Team tier adds authorized penetration testing and red-teaming, with applications limited to organizations. The Specialized tier has the fewest restrictions and is reserved for a small group authorized to test safety-critical systems, including power grids, flight systems and interbank transfer infrastructure. Anthropic vets each member with the US government, and existing Glasswing members will move into the Specialized tier.

Anthropic said on October 6 it is expanding its Cyber Verification Program, allowing vetted cybersecurity professionals to test its most powerful AI models with fewer safeguards. The company said its Project Glasswing initiative helped uncover more than 100,000 software vulnerabilities this year. Glasswing partners found at least 129,000 verified vulnerabilities between April and July, while Anthropic’s open-source scanning found 5,500 more between April and October. More than 33,000 vulnerabilities were rated critical or high severity. Anthropic said the figures likely undercount the total and estimated the true impact could be at least five times higher, because the data came from a survey of a limited number of partners.

The revamped program combines two programs Anthropic had operated for six months. Glasswing gave organizations protecting critical software access to Claude Mythos, Anthropic’s most cyber-capable model family; the earlier Cyber Verification Program gave vetted security teams reduced safeguards on Claude Opus and Sonnet. Anthropic’s April announcement of Claude Mythos Preview had raised fears that AI could hack software before it was secured.

The expanded program has three tiers, each with its own verification requirements and security controls, and all include Claude Opus 5.5, Sonnet 5.5, Mythos 5.1 and future models. The Defense tier covers incident response and malware analysis, and accepts applications from security teams, critical infrastructure operators, open-source maintainers and researchers with a record of reported vulnerabilities. The Red Team tier adds authorized penetration testing and red-teaming, with applications limited to organizations. The Specialized tier has the fewest restrictions and is reserved for a small group authorized to test safety-critical systems, including power grids, flight systems and interbank transfer infrastructure. Anthropic vets each member with the US government, and existing Glasswing members will move into the Specialized tier.