Breaking

Anthropic Moves Project Glasswing Participants Into Three-Tier Cyber Verification Program for Claude Models

October 7, 2026 at 00:46 EDT

Security
Foundation Models

Anthropic has expanded its Cyber Verification Program into a three-tier access system that brings Project Glasswing participants and newer Claude cyber capabilities under one verified framework. The program now offers qualifying security professionals tiered access to Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1 with reduced cyber blocking, while keeping separate safeguards for higher-risk uses.

Oct. 6, 2026 · Anthropic

Three Gates for Cyber Models

Anthropic expanded CVP into Defense, Red Team and Specialized tiers, moving Project Glasswing participants into a broader verified access system for Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1.

access tiers

129,000+

verified partner vulnerabilities

33,000+

critical or high severity

5,500

verified open-source findings

CyScenarioBench completed tasks

Red Team Access removed blocks in Anthropic's Opus 5.5 test

0/50

No CVP

4/50

Defense

34/50

Red Team

Each stacked block represents roughly four completed tasks; the Red Team column is proportional to the Defense result.

Defense Access

Incident response, malware analysis, vulnerability validation and maintainers with reported-vulnerability records.

→

Red Team Access

Authorized penetration testing by organizations, with blocks for physical harm and mass disruption.

→

Specialized Access

Fewest cyber blocks for verified organizations testing safety-critical systems.

Why defenders want it

Broader verified access could reduce blocks on legitimate malware analysis, exploit validation and large codebase review.

What still constrains it

Access remains vetted, monitored and tier-limited; teams still need human validation before fixes reach critical systems.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.