CVE-2026-65120

TensorRT-LLM SSRF can disclose information

NVIDIA · TensorRT-LLM

First documented disclosure · Review checkpoint

Summary

NVIDIA disclosed an additional server-side request-forgery flaw.

Severity and exploitation evidence

High · 7.1/10 · CVSS 3.1. Severity source.

Exploitation not established: The reviewed primary sources document a vulnerability or controlled demonstration, not confirmed malicious exploitation.

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

Unintended server-side requests cross a destination trust boundary; the bulletin does not identify the specific endpoint.

Conditions required

The vendor vector specifies local access without prior privileges or user interaction.

Potential impact

Potential information disclosure.

Affected versions and configurations

1.3.0rc24, all platforms, as listed by NVIDIA.

Fix and mitigation status

1.3.0rc25.

Practical next steps

Upgrade and limit inference-service access to internal and external network destinations.

Operating lesson

Limit outbound reach even when an inference service is not directly exposed to the internet.

Evidence and limits

Initial bulletin and CVE publication October 27. Separate from CVE-2026-24269; the public evidence does not justify merging them or claiming exploitation.

Sources