CVE-2026-24205

Concurrent TensorRT-LLM requests could cause data tampering

NVIDIA · NVIDIA TensorRT-LLM

First documented disclosure · Review checkpoint

Summary

Concurrent requests could trigger a race with data-integrity impact.

Severity and exploitation evidence

Medium · 6.4/10 · CVSS 3.1. Severity source.

Exploitation not established: Reviewed public sources describe the weakness, a correction or a controlled demonstration; malicious exploitation against users is not established.

Exploitation evidence source

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

Improper synchronization between concurrent requests.

Conditions required

A user can submit concurrent requests to an affected deployment.

Potential impact

Data tampering; further exploit details are not published.

Affected versions and configurations

All versions before 1.2.1, on all platforms.

Fix and mitigation status

1.2.1.

Practical next steps

Update TensorRT-LLM to 1.2.1 or later.

Operating lesson

Test request isolation under concurrency as part of inference-server security.

Evidence and limits

The current NVIDIA May bulletin includes this exact CVE, but its May 19/26 revision labels do not date the later individual entry. November 20 is the first verified issue-specific publication; no May disclosure date is inferred. NVIDIA assigns CVSS 3.1 6.4 Medium. The 1.2 fixes for four neighboring CVEs do not fix this race; it requires 1.2.1. CNA publication is 2026-11-20; this is distinct from the earliest reviewed public source and correction dates. nvidia assessments: CVSS 3.1 6.4. Full shared vendor bulletin reviewed: https://nvidia.custhelp.com/app/answers/detail/a_id/5805. Its page-level May date is not assigned to this later CVE entry.

Sources