analiza aiqbench

Anthropic is restructuring its Cyber Verification Program by integrating Project Glasswing into a three-tier framework that balances dual-use cybersecurity capabilities with conservative classifiers in general models. This expansion allows vetted defensive teams and authorized red teams to access advanced models like Claude Opus 5.5 while maintaining strict blocks on high-risk activities such as ransomware deployment. The move addresses the tension between providing powerful tools for critical infrastructure security and preventing misuse, offering a scalable path for organizations to leverage AI in cybersecurity operations.