The Information Machine

This story was merged into Anthropic's GLM-5.3 exploit capability report →

Merged·New·in the edition since 7 Oct 2026·5 sources·updated 7 Oct 2026

Anthropic's tiered cyber verification program

The gist

Anthropic Expands Cyber Verification Program to Three Access Tiers

The program gives vetted security professionals reduced AI safeguards for dual-use cybersecurity work, enabling offensive security tasks that Claude would otherwise block. Partner organizations have already used the program to identify over 129,000 verified software vulnerabilities.

The full picture

Anthropic expanded its Cyber Verification Program (CVP), merging Project Glasswing and its prior verification program into one structure with three tiers: Defense Access, Red Team Access, and Specialized Access. Defense Access is open to individuals and organizations; Red Team and Specialized Access are restricted to organizations only, and individuals must be on a paid plan to apply. Verified participants gain access to Claude Mythos 5.1, Opus 5.5, and Sonnet 5.5, including tiers that allow authorized offensive work such as penetration testing and red-teaming. The program unlocks high-risk dual-use activities but not outright prohibited uses; ransomware development and mass data exfiltration remain blocked regardless of verification status.

Eligibility factors include the nature of the organization's work, Anthropic's ability to verify identity, the legal and regulatory environment, and risk of diversion or compelled access. Anthropic said it takes a more cautious approach for organizations primarily serving military, intelligence, or law enforcement customers. Anthropic places each applicant at the highest tier the submitted information supports and aims to respond within seven business days.

Through Project Glasswing, partner organizations uncovered at least 129,000 verified software vulnerabilities between April and July 2026, with more than 33,000 rated critical- or high-severity. Several partners told Anthropic that the Claude Mythos models "had increased their rate of vulnerability finding by months or even years." Benchmark testing on CyScenarioBench showed that without CVP access every task was blocked, while in the Red Team Access tier no blocks occurred and Claude Opus 5.5 completed 34 of 50 tasks, matching its baseline 67.6% success rate.

Data retention is required for all participants to allow misuse monitoring; organizations on Zero Data Retention agreements are not currently eligible. An upcoming Enterprise Frontier Safeguards solution will eventually allow eligible organizations to store data in their own cloud infrastructure. The program is available on the Claude Platform, Google Cloud Vertex AI, and Microsoft Foundry, with Amazon Bedrock access limited to Enterprise Frontier Safeguards-eligible customers.

How it developed
First in the edition · 7 October 2026
6 October 2026

Anthropic launched an expanded Cyber Verification Program on October 6, merging Project Glasswing and its prior verification program into three tiers: Defense Access, Red Team Access, and Specialized Access.

Verified participants gain access to Claude Mythos 5.1, Opus 5.5, and Sonnet 5.5 for authorized offensive work such as penetration testing; ransomware development and mass data exfiltration remain blocked regardless of tier. Additional reporting the same day clarified that individuals may apply only for Defense Access and must be on a paid plan, organizations file a single application and are placed at the highest tier their submitted information supports, and Anthropic said it takes a more cautious approach for organizations primarily serving military, intelligence, or law enforcement customers.

Sources
The daily email

Want this in your inbox?

I send one email each morning with the stories that moved. If you would rather just read here, that works too.

Subscribe free