A critical authentication bypass vulnerability (CVE-2026-49468) has been disclosed in LiteLLM, an AI gateway that...

A critical authentication bypass vulnerability (CVE-2026-49468) has been disclosed in LiteLLM, an AI gateway that provides model access, fallbacks, and spend tracking across 100+ LLMs. The vulnerability, reported via OSV (GHSA-4xpc-pv4p-pm3w), allows an attacker to bypass authentication via host header injection. This could enable unauthorized access to virtual keys, budgets, and teams management, potentially compromising entire LLM deployments that rely on LiteLLM for access control. The security advisory was published on June 16, 2026, and affects all versions of LiteLLM prior to a patch. Given LiteLLM's role as a critical component in many AI infrastructure stacks, this vulnerability poses a significant security risk and may erode developer trust, leading to adoption delays or shifts to alternative gateways. The rapid release of multiple PyPI versions (1.84.9 through 1.89.2) around the same date suggests active development, but the lack of an immediate fix indicated in the advisory raises concerns. Source: OSV Advisory.