CVE-2026-4944

Published: Mag 28, 2026 Last Modified: Mag 28, 2026
ExploitDB:
Other exploit source:
Google Dorks:
HIGH 8,8
Source: [email protected]
Attack Vector: network
Attack Complexity: low
Privileges Required: none
User Interaction: required
Scope: unchanged
Confidentiality: high
Integrity: high
Availability: high

Description

AI Translation Available

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.

22

Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal')

Stable
Abstraction: Base Structure: Simple
Common Consequences
Security Scopes Affected:
Integrity Confidentiality Availability
Potential Impacts:
Execute Unauthorized Code Or Commands Modify Files Or Directories Read Files Or Directories Dos: Crash, Exit, Or Restart
Applicable Platforms
Technologies: AI/ML
Likelihood of Exploit: High
View CWE Details
https://huntr.com/bounties/97f706f7-a852-49b2-a4eb-76811e61…
https://huntr.com/bounties/97f706f7-a852-49b2-a4eb-76811e611daf