Silent Lynx Exploits WinRAR Flaw to Deploy Reverse Shell in CIS Government Attacks · 3h NGINX Fixes 5 Flaws, Including Possible RCE and a TLS Bug That May Skip Client-Cert Checks · 6h Next.js Ships Out-of-Band Patch for Two Critical RCE Flaws in Caching and SVG Processing · 6h Drupal Patches Critical Core RCE Flaw with No-Login Path via Contributed Modules · 7h Vulnerability in Sparx Systems Enterprise Architect software · 7h Curl verhelpt 'high' RCE-kwetsbaarheid gevonden door Anthropic · 7h Vulnerabilities in Slican telephone exchanges · 8h Palo Alto Networks waarschuwt voor misbruik van beveiligingslek in firewalls · 11h Cisco meldt actief misbruik van kritiek lek in Secure Email Gateway · 11h Langflow's Critical Authenticated RCE Flaw Targeted in 1,423 Honeypot Exploit Attempts · 12h Silent Lynx Exploits WinRAR Flaw to Deploy Reverse Shell in CIS Government Attacks · 3h NGINX Fixes 5 Flaws, Including Possible RCE and a TLS Bug That May Skip Client-Cert Checks · 6h Next.js Ships Out-of-Band Patch for Two Critical RCE Flaws in Caching and SVG Processing · 6h Drupal Patches Critical Core RCE Flaw with No-Login Path via Contributed Modules · 7h Vulnerability in Sparx Systems Enterprise Architect software · 7h Curl verhelpt 'high' RCE-kwetsbaarheid gevonden door Anthropic · 7h Vulnerabilities in Slican telephone exchanges · 8h Palo Alto Networks waarschuwt voor misbruik van beveiligingslek in firewalls · 11h Cisco meldt actief misbruik van kritiek lek in Secure Email Gateway · 11h Langflow's Critical Authenticated RCE Flaw Targeted in 1,423 Honeypot Exploit Attempts · 12h
🛡️
ACyber Sentinel Security Assessment Team
← Back to feed
⚠️ Vulnerability Report

CVE-2026-4944

CVE-2026-4944 - Hardcoded trust_remote_code=True in vllm-project/vllm Bypasses User Security Control

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.

Published: May 28, 2026 19:16

Last Modified: Jun 17, 2026 10:57