Fracturing Software Security With Frontier AI Models

Unit 42 warns that frontier AI models materially change the threat landscape by enabling rapid, automated discovery and exploitation of vulnerabilities—particularly in open-source software—leading to faster, higher-scale attacks across the full kill chain. The report details AI-enabled reconnaissance, autonomous lateral movement, custom exploit generation, and LLM-driven analysis of exfiltrated data, and prescribes defensive actions including assumed-breach operations, SBOMs, hardened build environments, accelerated patching, and automated incident response.