CVE-2026-24142 is a deserialization flaw (CVSS 6.3, medium) in NVIDIA TensorRT-LLM on any platform. An unsafe serialized handle could lead to code execution, data tampering and information disclosure. NVIDIA's May 2026 bulletin lists versions before v1.2 as affected and recommends updating to v1.2.1.
What's affected
TensorRT-LLM versions before v1.2, per NVIDIA's May 2026 security bulletin. The bulletin recommends cloning or updating to v1.2.1 from GitHub, which also covers other CVEs in the same bulletin. The CVSS vector in the record is local access with low privileges, so the risk grows on shared machines.
How to check if you're exposed
pip show tensorrt_llm
Check each environment that runs inference, including containers built from NVIDIA base images, and note whether the host is shared by several users or services. Multi-GPU deployments are called out in the bulletin's mitigation text.
What to do now
- Update TensorRT-LLM to v1.2.1 or later.
- Read the bulletin's mitigation section. It has guidance for multi-GPU deployments that mentions the trtllm-llmapi-launch executable.
- Rebuild and redeploy containers that include the library.
- Limit local access on inference hosts to the accounts that need it.
Why it matters
The weakness is CWE-502, deserialization of untrusted data. Serialization formats that rebuild objects can run code while reading data, so any serialized handle exchanged between processes should come from a trusted peer. The medium score reflects the local attack vector and conditions required. The record does not list this CVE as known exploited.
Reading the severity score
The record gives a CVSS base score of 6.3, published 2026-05-20. In the vector, the attacker works with local access, needs a low privilege account, and the attack is low in complexity. User interaction is not needed. Impact on confidentiality is low, on integrity low and on availability low.
A base score describes the flaw in general, not your risk. Weigh it against how exposed your instance is and what it can reach. The Gold page for the CVE keeps the score, weakness and references together.
Find it across your own code
To see which of your packages are affected, the free in-browser dependency scan at gold.safeguard.sh/scan takes a manifest and lists known vulnerabilities. Reachability analysis then helps you judge whether your code calls the vulnerable path, so you can order the work.
Try it yourself
Click through it at your own pace, or open it full screen.
Quick check
Why is a local flaw worth patching?
Inference servers are often shared by several services or users, and a local attacker or a compromised neighbor process can then reach the vulnerable handling.
Sources
Last reviewed: 19 October 2026.
