CVE-2026-47155
A flaw was found in vLLM, an inference and serving engine for large language models LLMs. The revision pinning controls in vLLM do not consistently apply to all artifacts loaded for a model. This allows a deployment configured with specific revisions to still load dynamic code or other...