Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

NewsPolicy

Anthropic opens Claude to vetted cyber teams in three tiers, and Glasswing partners report 129,000 verified flaws

Anthropic expanded its Cyber Verification Program into three tiers with fewer cyber safeguards on Claude Opus 5.5, Sonnet 5.5 and Mythos 5.1. Defense Access covers SOC work, incident response, malware reverse engineering and vulnerability analysis, for security teams at companies, nonprofits, universities and government bodies, infrastructure operators and open-source maintainers, with answers in a few days. Red Team Access adds authorized penetration testing for organizations only, after a review of a few weeks; actions such as deploying ransomware stay blocked. Specialized Access, the most permissive, is for testing systems such as power grids and telecom networks, and every organization is reviewed with the US government; Project Glasswing members move there. Anthropic says Glasswing partners found 129,000 verified vulnerabilities from April to July and its own open-source scanning 5,500 more, over 33,000 of them critical or high. On its CyScenarioBench test, a Defense Access account had 46 of 50 offensive tasks blocked; Red Team Access had none blocked and Claude Opus 5.5 completed 34 of 50.

Why it matters

The most capable models for finding and exploiting bugs are now gated by who you are, not only by what you ask. Defenders who qualify should apply; everyone else should expect the flood of AI-found vulnerabilities, and the patches, to keep growing.

Sources

Read it with the rest of the news in the Learning Center