Anthropic expands Claude cyber access with tiered verification
Anthropic has broadened its Cyber Verification Program by organizing access into three tiers based on intended use. The revised safeguards are more flexible for some users, but the company is not extending the same access to everyone. The program controls how Claude can be used for cybersecurity-related tasks.
Anthropic’s revised approach to its Cyber Verification Program places access to Claude for cybersecurity work into three categories, each tied to what the user intends to do. The company describes the updated safeguards as more adaptable for certain users, while still withholding the same level of access from others. The program therefore acts as a gatekeeping mechanism for how Claude may be applied to security-focused tasks. This fits a broader pattern in AI: providers are trying to distinguish legitimate defensive or research uses from potentially harmful ones, often through verification and tiered permissions rather than a single blanket rule.
Security researchers, developers, and organizations that rely on AI for defensive cybersecurity could be affected by how Anthropic assigns access. Tiered verification may make advanced assistance available to some approved users while leaving others with fewer capabilities, potentially influencing how quickly vulnerabilities are investigated or incidents are handled. It may also shape expectations around transparency, accountability, and equitable access to powerful AI tools. The practical effect will depend on how clearly the tiers are defined and how consistently they are applied.