Anthropic Expands Claude Access for Security Teams with Reduced Safety Filters

Anthropic has expanded access to its AI model Claude for security teams, allowing vetted organizations to utilize the model with fewer safety restrictions. This initiative, part of the Cyber Verification Program, aims to facilitate vulnerability research, malware analysis, incident response, and penetration testing. Manuel Uth stated that the program is open to corporate security teams, government agencies, universities, critical infrastructure operators, open-source developers, and individual researchers.

Between April and July 2026, partners confirmed 129,000 vulnerabilities, with 33,000 classified as high-severity or critical. The real-world impact of these vulnerabilities is estimated to be five times higher than reported. The program includes various access levels: Defense Access for general use, Red Team Access for authorized attack simulations limited to organizations, and Specialized Access for critical system testing, such as flight controls and banking infrastructure.

Several partners reported that Claude significantly accelerated their work, reducing timelines by months or even years. This follows Anthropic's previous initiatives to enhance security capabilities, including the launch of Project Glasswing. The expanded access is expected to empower security teams to better address the growing number of vulnerabilities in their systems.