CVE-2026-4944

What it is

vllm-project/vllm version 0.14.1 contains a vulnerability where the `trust_remote_code=True` parameter is hardcoded in two model implementation files (`vllm/model_executor/models/nemotron_vl.py` and `vllm/model_executor/models/kimi_k25.py`). This bypasses the user's explicit `--trust-remote-code=False` setting, enabling remote code execution via malicious HuggingFace model repositories. This issue is an incomplete fix for CVE-2025-66448 and CVE-2026-22807, as it affects separate code paths in model implementation files. Deployments loading NemotronVL or KimiK25 models are particularly impacted.

At a glance

Exploitability

Can be exploited over the network without logging in, but needs a user to act. High risk of data exposure, data tampering, service disruption.

Exploit odds

0.7% chance of exploitation in the next 30 days (EPSS).

Fix

No patch or advisory is linked yet. Look at mitigations or restrict exposure.

Timeline

1. 2026-05-28 19:16 UTC PublishedDisclosed publicly on CVE.org and NVD
2. 2026-06-17 10:57 UTC Last updatedMost recent change to the record

Details

Published

2026-05-28 19:16 UTC

Last modified

2026-06-17 10:57 UTC

Exploit probability

0.7% · EPSS, next 30 days

Weakness

CWE-22 · Path traversal

Record status

Deferred

Attack profile CVSS 3.0 · 8.8

Attack vector

Network

Complexity

Low

Privileges needed

None

User interaction

Required

Scope

Unchanged

Confidentiality

High

Integrity

High

Availability

High