Anthropic reported that its frontier model, Claude Opus 4.6, autonomously identified more than 500 previously unknown zero‑day vulnerabilities across open‑source libraries during internal red‑team testing. The company framed the capability as a defender’s tool but acknowledged the 'dual‑use' risk: attackers could weaponize the same models to find and exploit flaws before patches are available. Anthropic’s Frontier Red Team and Logan Graham, head of the team, said the findings show large language models can augment vulnerability discovery, but the company is deploying probes and monitoring to detect misuse. The development sharpens a dilemma for campus cybersecurity teams: AI can accelerate discovery and remediation, but also raises the speed and scale of potential exploit discovery. Universities running critical research infrastructure, open‑source projects and institutional apps must reassess vulnerability disclosure policies, accelerate patch management, and coordinate with industry partners to ensure defenders gain prioritized access to AI‑assisted security tools.