Anthropic expands cyber verification programme into three tiers

Anthropic said it had launched an updated Cyber Verification Program that folded in Project Glasswing. The programme was organised into Defense Access, Red Team Access and Specialized Access, each with different verification requirements.

Defense Access covered defensive work such as incident response, malware reverse engineering and vulnerability analysis. Red Team Access added authorised penetration testing, while Specialized Access was limited to verified organisations testing systems that could affect people, markets or critical services.

Anthropic said it reviewed the most restricted tier with the US government. It said current Project Glasswing members would move to that tier without reapproval for their existing models.

The company said the programme made Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1 available under the tiers, while its general models kept stronger cyber blocks. It also said Enterprise Frontier Safeguards would later combine zero data retention with safeguards, and that some users of Claude Fable 5.1 or Claude Mythos 5.1 could use CVP with zero retention until then.

Anthropic said testing on CyScenarioBench showed different results by tier, with all tasks blocked immediately without CVP and 46 of 50 trials blocked in Defense Access. It said Red Team Access allowed Claude Opus 5.5 to complete 34 of 50 tasks, and that Project Glasswing partners had found at least 129,000 verified vulnerabilities between April and July 2026. Anthropic said its own open-source scanning found another 5,500 between April and October 2026, and that more than 33,000 of the verified vulnerabilities were rated critical or high severity.