Anthropic Unifies Cyber Verification and Glasswing Programs
Anthropic announced on Tuesday the launch of a revised Cyber Verification Program (CVP) that integrates the former CVP and Project Glasswing into a single, three-tier access framework for its most advanced AI models. The update streamlines access to Claude Opus 5.5, Sonnet 5.5, Mythos 5.1, and future models under a unified structure, replacing the previous dual-program approach.
Prior to this change, Glasswing provided access to Claude Mythos for organizations securing critical software, while the original CVP permitted approved teams to use Opus and Sonnet models with reduced safeguards. Now, all three tiers include access to Opus 5.5, Sonnet 5.5, Mythos 5.1, and upcoming models, each with distinct verification requirements and security controls tailored to the user’s role and risk profile.
Defense Access Tier for Protective Security Work
The Defense Access tier supports security operations center (SOC) activities, incident response, malware reverse engineering, and vulnerability analysis and validation. Eligible applicants include internal security teams defending their own systems, critical infrastructure operators, small security firms, open source maintainers, and individual researchers with a documented history of reporting vulnerabilities. Anthropic states it aims to process these applications within a few days.
Red Team Access for Authorized Penetration Testing
The Red Team Access tier adds permissions for authorized penetration testing and red teaming, strictly limited to systems the organization is explicitly permitted to test. Anthropic emphasizes that actions posing risks of physical harm or mass disruption—such as deploying ransomware—are blocked in real time, regardless of tier. Currently, this tier is restricted to organizations only; individual researchers are not eligible. Reviews for Red Team Access are expected to take several weeks, though qualifying applicants receive Defense Access privileges during the evaluation period.
Specialized Access for Safety-Critical System Testing
The Specialized Access tier imposes the fewest cyber safeguards and is reserved for a limited number of organizations authorized to test safety-critical systems, including power grids, flight control systems, telecom networks, and interbank transfer infrastructure. Anthropic conducts vetting for this tier in collaboration with the U.S. government. Existing Glasswing participants are being transitioned into this tier as part of the program update.
According to Anthropic, Glasswing partners identified at least 129,000 verified vulnerabilities between April and July, while Anthropic’s own open source scanning uncovered an additional 5,500 vulnerabilities between April and October. Of these, more than 33,000 were rated critical or high severity. The company notes that the actual impact is likely at least five times greater, as the reported figures represent only a subset of Glasswing participants.
Data Retention, Deployment Options, and Transition for Existing Users
All organizations participating in the new program must agree to data retention policies that allow Anthropic to monitor for misuse. Until the fall release of Enterprise Frontier Safeguards—which will permit eligible customers to store data in cloud infrastructure they control—organizations requiring zero data retention can access Fable 5.1 or Mythos 5.1 through CVP with no data retention.
The program is available on the Claude Platform, Google Cloud’s Vertex AI, and Microsoft Foundry. On Amazon Bedrock, access is limited to customers eligible for Enterprise Frontier Safeguards. Existing CVP members will retain their current configurations for prior models and will be automatically assessed for eligibility to access the new models under the updated framework.
Source: SecurityWeek