Anthropic has announced the expansion of its Cyber Verification Program (CVP), building on the success of Project Glasswing, which identified over 129,000 software vulnerabilities in just four months. The initiative provides vetted cybersecurity professionals with reduced-safeguard access to advanced AI models like Claude Opus 5.5, enabling specialized tasks in defense and red-teaming that would otherwise be blocked by standard safety filters.
Despite the surge in AI-driven vulnerability discovery, research indicates that actual exploitation remains low, with only a tiny fraction of Anthropic-identified flaws currently observed in the wild. Furthermore, security experts warn that AI-generated code still presents significant risks, as nearly 44% of AI-produced patches or code snippets may introduce new vulnerabilities, emphasizing the ongoing need for human oversight in AI-assisted security workflows.
Top comments (0)