Koffei

CVE-2026-42271

🔴 High In CISA KEV

LiteLLM (AI/MCP gateway, BerriAI); Red Hat OpenShift AI · ≥ 1.74.2 and < 1.83.7; OpenShift AI 2.25 before 2.25.8, 3.3 before 3.3.4, 3.4

CVSS (Koffei note)
8.8
CVSS (v4.0, CVE List)
8.7
EPSS (2026-11-18)
91.445%
99.82th percentile
CISA KEV
2026-06-08
due 2026-06-22

Exploitation

Yes — CISA KEV since 08/06/2026 (due 22/06/2026; CISA disclosed neither actors nor method). Most recent dated exploitation: SANS ISC honeypots, 14-15/11/2026, POST /mcp-rest/test/tools/list running id, chained with CVE-2026-48710 through malformed Host headers (diary of 16/11/2026). Earlier: Microsoft (26/08/2026) reported the chain used to reach LiteLLM gateways and drop XMRig; Wiz (27/08/2026) names this CVE in a LiteLLM/MCP cryptomining campaign. Rated 🔴: KEV plus exploitation evidence dated 15/11/2026, inside the 60-day window (tested 17/11/2026). A Qilin ransomware link (CybelAngel; Wiz citing "external researchers") has no primary source and is unverified; the KEV ransomware flag was not read. Amazon's ARTEX report of 13/11/2026 names LiteLLM targets but not this CVE. A public PoC is referenced by SANS (not opened; sploitus not checked). cyberveille RSS: checked 17/11/2026 (feed up to 16/11/2026), no result

Timeline

  1. 08/05/2026CVE record published (CVE List)
  2. 08/06/2026Added to the CISA KEV catalogue
  3. 22/06/2026CISA deadline for US federal agencies
  4. 17/11/2026Fresh scans for the LiteLLM and Starlette pre-auth RCE chain

Relationships

CVE-2026-20316CVE-2026-20316CVE-2026-20079CVE-2026-20079CVE-2026-0257CVE-2026-0257CVE-2024-21762CVE-2024-21762CVE-2023-27532CVE-2023-27532LiteLLMLiteLLMQilinQilinCVE-2026-42271CVE-2026-42271

exploits suspected named in the same story vendor · Select a box to open its page · All relationships

Risk

8.8 High (v3.1) — AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H; 8.7 High (v4.0, GHSA) — AV:N/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:H/SC:H/SI:N/SA:N, EPSS 91.445% (OpenCVE and vulmon agree) (CWE-77/CWE-78: the MCP preview endpoints POST /mcp-rest/test/connection and /mcp-rest/test/tools/list spawn whatever stdio command the request supplies as a subprocess on the proxy host, gated only by a valid proxy API key with no role check)

Fix

Yes — 1.83.7 (released 08/05/2026), with Starlette ≥ 1.0.1 to close the unauthenticated chain; Red Hat RHSA-2026:28960 (2.25), RHSA-2026:30056 (3.3), RHSA-2026:27784 (3.4)

Official description

LiteLLM is a proxy server (AI Gateway) to call LLM APIs in OpenAI (or native) format. From version 1.74.2 to before version 1.83.7, two endpoints used to preview an MCP server before saving it — POST /mcp-rest/test/connection and POST /mcp-rest/test/tools/list — accepted a full server configuration in the request body, including the command, args, and env fields used by the stdio transport. When called with a stdio configuration, the endpoints attempted to connect, which spawned the supplied command as a subprocess on the proxy host with the privileges of the proxy process. The endpoints were gated only by a valid proxy API key, with no role check. Any authenticated user — including holders of low-privilege internal-user keys — could therefore run arbitrary commands on the host. This issue has been patched in version 1.83.7.

CVSS:4.0/AV:N/AC:L/AT:P/PR:L/UI:N/VC:H/VI:H/VA:H/SC:H/SI:N/SA:N

CWE-77,CWE-78