CompanyClaude

Anthropic reduces model blocks for security pros

Original title: Expanding the Cyber Verification Program

Plain language

What it is: Anthropic is launching an expanded version of its Cyber Verification Program. This program gives approved security professionals access to advanced cyber capabilities with fewer automated blocks on the models.
How it works: The program now has three levels of access called Defense Access, Red Team Access, and Specialized Access. Security teams must apply, pass verification, and meet specific security controls to get access to models like Claude Opus 5.5 and Claude Sonnet 5.5.
Where it is available: The program is available on the Claude Platform, Google Cloud's Vertex AI, and Microsoft Foundry. It is also available on Amazon Bedrock for customers who qualify for Enterprise Frontier Safeguards.
Example: A security team at a company or a regional hospital can use the Defense Access tier to analyze vulnerabilities and respond to security incidents.

What it is: Anthropic has expanded its Cyber Verification Program (CVP) by merging it with Project Glasswing and introducing three distinct access tiers. This allows verified security researchers and teams to use Claude models with reduced blocking classifiers.
How it works: Users apply for a specific tier, such as Defense Access or Red Team Access, which determines the level of safety classifier relaxation. Admins then assign this access to specific workspaces, and data retention is required unless using Enterprise Frontier Safeguards or specific zero-retention models.
Where it is available: The program is accessible via the Claude Platform, Google Cloud's Vertex AI, and Microsoft Foundry, as well as Amazon Bedrock for eligible Enterprise Frontier Safeguards users.
Example: An authorized red-teaming firm can use the Red Team Access tier to run penetration tests on IT systems without triggering the standard real-time safety blocks.

The headline and explanation on this page are based on Anthropic's post. The full details are in the original.

Expanding the Cyber Verification Program