AI models generate active software security exploits and identify system flaws

OpenAI demonstrated an automated system that scanned over two thousand potential Linux kernel vulnerabilities and produced working exploit scripts alongside proposed patches. In a related update, Anthropic's Claude assistant received official attribution in a security advisory for uncovering memory exposure bugs within the Xen hypervisor.