#1 Nicholas Carlini: AI Now Capable of Autonomous Zero-Day Vulnerability Exploitation
At the [un]prompted 2026 security conference, Anthropic research scientist Nicholas Carlini demonstrated that language models can now autonomously identify and exploit zero-day vulnerabilities. His presentation highlighted that even software as extensively audited as the Linux kernel is susceptible to these AI-driven attacks. This development marks a significant shift in cybersecurity, raising urgent questions about the future of software security and how the industry can defend against increasingly capable autonomous exploitation tools.