Anthropic Expands Cybersecurity Program to Regional Hospitals & Critical Infrastructure

The initiative aims to help organizations strengthen cybersecurity while limiting the misuse of AI tools for malicious activities.

Anthropic has expanded its Cyber Verification Program (CVP) to provide vetted security professionals, including teams at regional hospitals and critical infrastructure organizations, with access to advanced AI capabilities for identifying vulnerabilities, testing defenses, and responding to cyber threats.

The redesigned program introduces three access tiers Defense Access, Red Team Access, and Specialized Access with different eligibility requirements and security controls. The initiative aims to help organizations strengthen cybersecurity while limiting the misuse of AI tools for malicious activities.

Anthropic said its AI models have dual-use capabilities that can help security teams identify and fix vulnerabilities but could also enable attackers to exploit them. Its generally available models therefore retain conservative cybersecurity safeguards that restrict many advanced security-related tasks.

Three Tiers for Cybersecurity Access

The Defense Access tier targets security teams focused on protecting organizational IT systems, critical infrastructure, and smaller security firms. Eligible users can access capabilities for vulnerability assessment, incident response, and malware reverse engineering. Regional hospitals and municipal utilities are among the organizations Anthropic identifies as potential participants.

The Red Team Access tier supports authorized penetration testing and adversarial security assessments. Participants must meet higher eligibility requirements and comply with additional safeguards. Anthropic said the system will continue blocking actions that could cause physical harm or widespread disruption, including ransomware deployment and testing against high-risk safety systems.

The Specialized Access tier is reserved for a limited number of verified organizations authorized to test safety-critical systems. These include flight operating systems, power grids, telecommunications networks, interbank transfer infrastructure, and government administrative networks. US government involvement will be required to approve participants. Organizations currently participating in Anthropic's Project Glasswing will transition to this tier.

All three tiers provide access to Claude models, including Opus 5.5, Sonnet 5.5, and Mythos 5.1, according to the announcement.

Organizations interested in participating can apply through Anthropic's Cyber Verification Program portal. The company also said data retention is mandatory for enrolled organizations to support monitoring for potential cybersecurity misuse.

Healthcare Cybersecurity in Focus

The expansion follows Anthropic's Project Glasswing initiative, launched in April 2026, which brought together governments and approximately 150 selected organizations from critical sectors, including healthcare, to identify and test software vulnerabilities using advanced AI capabilities.

Electronic health record provider Epic participated in the initiative and identified security vulnerabilities that prompted a temporary pause in some product development activities while fixes were implemented.

The broader cybersecurity push comes as AI companies warn that increasingly capable models could make cyberattacks more sophisticated. Anthropic also joined a collective cybersecurity initiative led by OpenAI, calling on organizations responsible for critical infrastructure, including hospitals and water treatment facilities, to strengthen their defenses.

Anthropic said its expanded program consolidates access previously offered through Project Glasswing and the Cyber Verification Program, allowing more vetted security organizations to use advanced AI tools for defensive work.

Stay tuned for more such updates on Digital Health News