GovernanceAnnouncement

Anthropic expands cybersecurity verification tiers

Anthropic is expanding its Cyber Verification Program into three access tiers. Qualifying security professionals can apply for capabilities matched to their work, with different verification and security requirements.

Authorization remains part of access

The program announcement brings together earlier verification and Project Glasswing efforts. Defense Access covers incident response, malware analysis and vulnerability validation. Red Team Access adds authorized penetration testing and red-teaming. Specialized Access permits narrowly reviewed, authorized high-risk work for safety testers, with additional verification and security requirements.

Anthropic added three tiers to its Cyber Verification Program for qualifying security professionals and organizations. Verified access supports defensive and authorized testing work with fewer inappropriate refusals, while eligibility and the scope of permitted work still apply.
Trion

Three access tiers, distinct limits

Defense Access
Authorized scopeIncident response, malware and vulnerability checks
BoundaryDefined defensive scope
Red Team Access
Authorized scopeAuthorized penetration testing
BoundaryPhysical harm and major disruption blocked
Specialized Access
Authorized scopeNarrowly reviewed high-risk safety work
BoundaryAdditional verification and security requirements
View data
TierAuthorized scopeBoundary
Defense AccessIncident response, malware and vulnerability checksDefined defensive scope
Red Team AccessAuthorized penetration testingPhysical harm and major disruption blocked
Specialized AccessNarrowly reviewed high-risk safety workAdditional verification and security requirements

Anthropic · published 2026-10-06. Source-bound illustration, not a performance benchmark.

Download image

Anthropic says generally available models retain conservative cyber safeguards, while verification reduces inappropriate blocking for qualifying work. Red Team Access still blocks actions with potential for physical harm or major disruption. Each tier has a defined scope; eligibility is not blanket permission to test any system.

Original sources

  1. Anthropic: original announcementwww.anthropic.com

Checked 9 Oct 2026 · A manually curated edition. Availability may change; company performance claims are not Trion test results. Editorial method.