Claude AI Discovers 500+ Zero-Day Vulnerabilities in Major Open-Source Libraries

Anthropic's Claude Opus 4.6 AI discovers over 500 previously unknown zero-day vulnerabilities in major open-source libraries including GhostScript, OpenSC, and CGIF, prompting new security controls to prevent malicious exploitation of the AI's powerful cyber capabilities.

• Anthropic's Claude Opus 4.6 discovers over 500 previously unknown zero-day vulnerabilities in open-source libraries using only out-of-the-box capabilities without specialized instructions
• The AI model finds critical security flaws in popular utilities including GhostScript, OpenSC, and CGIF that could crash systems or corrupt memory, with each vulnerability validated by security researchers
• Anthropic implements new security controls to prevent malicious abuse of Claude's cyber capabilities while exploring ways to bring vulnerability detection tools to the broader cybersecurity community