Beginning the weekend of November 15–16, 2026, threat actors initiated active scanning of internet-facing infrastructure for vulnerabilities in LiteLLM and Starlette. The exploitation involves a chained attack using CVE-2026-42271 and CVE-2026-48710.

CVE-2026-42271 is an unauthenticated remote code execution vulnerability in LiteLLM's /mcp-rest/test/tools/list endpoint. CVE-2026-48710, known as 'BadHost', is a host-header validation flaw in Starlette, allowing attackers to manipulate the Host header.

Honeypot telemetry confirmed active, automated exploitation attempts using a Python aiohttp user-agent. This indicates that attackers are employing automated tools to exploit these vulnerabilities.

Organizations using unpatched versions of LiteLLM or Starlette prior to 1.0.1 are at immediate risk of full system compromise. Attackers have been performing automated reconnaissance and scanning for the vulnerable endpoint in LiteLLM.

The vulnerabilities were disclosed as early as July 2026 and have been continuously re-weaponized by attackers. The User-Agent 'Python/3.14 aiohttp/3.14.5' has been identified in the exploit attempts, suggesting the use of automated exploit tooling.