Millions of AI agents imperiled by critical vulnerability in open source package
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A critical vulnerability dubbed 'BadHost' (CVE-2026-48710) has been discovered in Starlette, a Python ASGI framework with 325 million weekly downloads. The flaw allows attackers to bypass path-based authorization by injecting a single character into the HTTP Host header. Because Starlette underpins FastAPI and many AI-related packages — including vLLM, LiteLLM, MCP servers, and agent harnesses — millions of AI agents and services are exposed. Attackers can breach servers and steal credentials stored by MCP servers. The vulnerability is trivially exploitable against systems not behind a properly configured firewall. A fix is available in Starlette 1.0.1, and an online scanner has been released to check server exposure.
Sort: