Anthropic has unified its Cyber Verification Program and Project Glasswing into a three-tier access model for its most capable AI systems, expanding controlled use of Claude models for defensive security work. The program is designed to support vulnerability research, incident response, and authorized testing while keeping dangerous actions like ransomware deployment blocked, and it has already helped uncover tens of thousands of vulnerabilities. #Anthropic #ClaudeOpus5.5 #ClaudeSonnet5.5 #ClaudeMythos5.1 #ProjectGlasswing
Keypoints
- Anthropic merged CVP and Project Glasswing into one revamped program.
- The new structure includes Defense Access, Red Team Access, and Specialized Access.
- Eligible users can study malware, analyze vulnerabilities, and perform authorized red teaming.
- Ransomware and other harmful actions remain blocked in real time.
- Glasswing partners and Anthropic found more than 134,000 verified vulnerabilities.
Read More: https://www.securityweek.com/anthropic-introduces-3-tier-cyber-verification-program-for-ai-access/