vLLM

vLLM is an open-source library for running large language models efficiently. This high-severity vulnerability requires local access with no authentication or user privileges needed, but does require user interaction, and could result in complete compromise of system confidentiality, integrity, and availability. Security teams should monitor for exploitation attempts targeting local LLM deployments and consider implementing access controls on systems running vLLM until patches are available after the July 2026 deadline.