CVE-2026-73555

A flaw was found in vLLM, an inference and serving engine for large language models. Unauthenticated attackers can send malformed JSON requests to various API endpoints, triggering validation errors. The system's error handling mechanism converts these validation errors into strings without...