Anthropic Expands Cyber Verification Program, Empowering Security Teams with Advanced AI Models

Anthropic has announced a significant expansion of its Cyber Verification Program (CVP), providing vetted cybersecurity professionals with tiered access to its advanced AI models, including Claude Opus 5.5, Claude Sonnet 5.5, and Claude Mythos 5.1. This program is designed to allow security teams to utilize these powerful AI tools with reduced safeguards and blocking classifiers, enabling more effective cybersecurity operations. The expansion builds upon the success of Project Glasswing, which, between April and July 2026, identified at least 129,000 verified software vulnerabilities, with an additional 5,500 found through open-source scanning efforts by October 2026. This initiative is critical for the cybersecurity community. As AI capabilities rapidly advance, so does the potential for AI-powered attacks. By providing defenders with access to similarly advanced AI tools, Anthropic is helping to level the playing field. The tiered access — Defense Access, Red Team Access, and Specialized Access — allows organizations to choose the level of model access that best suits their specific security needs, from incident response and malware analysis to authorized penetration testing. This directly impacts practitioners by giving them sophisticated tools to enhance their defensive posture, automate tedious tasks, and potentially uncover vulnerabilities that human analysts might miss. The expansion of the CVP fits within a broader trend of leveraging AI for enhanced security. As software complexity grows and the attack surface expands, traditional security methods struggle to keep pace. Generative AI models, with their ability to analyze vast amounts of data, identify patterns, and even generate code, are becoming indispensable for proactive threat detection, vulnerability management, and automated response. This move by Anthropic aligns with the industry's increasing reliance on AI to augment human security expertise and build more resilient systems. Other companies are also exploring AI's role in security, from AI-powered fuzzing to using AI for secret protection in codebases. In practice, this means cybersecurity teams should evaluate how these advanced AI models can be integrated into their existing workflows. Practitioners should consider applying for the CVP to explore how Claude's capabilities can enhance their vulnerability assessments, incident response playbooks, and red teaming exercises. It also highlights the growing importance of AI literacy for security professionals, as understanding how to effectively leverage and interpret AI outputs will be crucial. Organizations will need to establish clear guidelines and ethical frameworks for using such powerful AI tools, especially when operating with reduced safeguards, to ensure responsible and effective deployment.