Anthropic Expanding the Cyber Verification Program

Anthropic has launched an expanded Cyber Verification Program (CVP) to provide vetted security professionals with tiered access to its most capable models—including Claude Opus 5.5, Claude Sonnet 5.5, and Claude Mythos 5.1—with reduced blocking classifiers. This program aims to empower defenders to identify and fix vulnerabilities more effectively while maintaining safeguards to prevent malicious exploitation.

Tiered Access for Security Professionals

Anthropic has implemented three distinct access tiers based on the scope of the user's cybersecurity work, each with its own verification requirements and security controls.

Defense Access

Defense Access is designed for defensive cybersecurity tasks such as incident response, malware reverse-engineering, and vulnerability analysis. Qualifying entities include:

  • Security teams at companies, nonprofits, universities, and government bodies defending their own systems.
  • Operators of critical infrastructure (e.g., regional hospitals, municipal utilities).
  • Small security firms and open-source maintainers.
  • Individual researchers with a documented track record of reporting vulnerabilities.

Red Team Access

Red Team Access expands upon Defense Access by allowing authorized penetration testing and red-teaming. This tier is available only to organizations, not individual researchers. While it allows for more offensive capabilities, real-time blocks remain in place for actions that could cause mass disruption or physical harm, such as deploying ransomware or testing high-risk safety systems.

Specialized Access

Specialized Access provides the fewest cyber blocks and is reserved for a limited set of verified organizations authorized to test critical safety systems that impact lives or markets, such as power grids, telecom networks, and flight operating systems. Access to this tier is reviewed in depth in collaboration with the US government; existing members of Project Glasswing transition into this tier.

Efficacy and Performance Benchmarks

To validate the tiered safeguards, Anthropic tested Claude Opus 5.5 using CyScenarioBench, an evaluation measuring the ability to plan and execute multi-stage cyber operations. The results demonstrate a clear distinction in capability access across tiers:

  • Generally Available Models: Every task was blocked on the first prompt.
  • Defense Access Tier: 46 out of 50 trials were blocked at some point during the challenge, with only four tasks succeeding.
  • Red Team Access Tier: No blocks occurred, and the model successfully completed 34 of 50 tasks (a 68% success rate), which is effectively equivalent to the model's performance with no safeguards applied.

Impact on Vulnerability Discovery

Data from Project Glasswing indicates that advanced models, specifically Claude Mythos, significantly accelerate the identification of software vulnerabilities. Between April and July 2026, partners uncovered at least 129,000 verified software vulnerabilities. Additionally, Anthropic's own open-source scanning found 5,500 verified vulnerabilities between April and October 2026.

Of these combined verified vulnerabilities, more than 33,000 were rated as critical- or high-severity. Anthropic notes that these figures are likely a lower bound, as they are based on partial data from 33 partner reports and a subset of partners who disclosed patched numbers.

Data Retention and Privacy

Currently, data retention is required for organizations in the CVP to monitor for cyber misuse. However, Anthropic is developing "Enterprise Frontier Safeguards" (EFS), a solution combining zero data retention with robust safeguards, expected later this fall. Until EFS is released, organizations using Claude Fable 5.1 or Claude Mythos 5.1 with zero data retention can continue to use CVP under those terms.

Availability and Application

The Cyber Verification Program is available via the Claude Platform, Google Cloud's Vertex AI, and Microsoft Foundry. On Amazon Bedrock, it is available only to customers eligible for Enterprise Frontier Safeguards.

Organizations can apply for access through the Anthropic portal, where they will be required to provide proof of security controls and undergo verification.

Sources