GPU VulnDB

Database/AI/ML frameworks & serving

NVIDIA Model Optimizer: deserialization of untrusted data can lead to code execution

CVSS 8.8CVE-2026-67455AI/ML frameworks & servingcurated

Impact

NVIDIA Model Optimizer deserializes untrusted data. NVIDIA says a successful exploit might lead to code execution, data tampering, denial of service and information disclosure. Model Optimizer runs inside quantization and optimization jobs on GPU nodes, so code execution there runs with that job's access to the node's GPUs, model weights, datasets and any credentials mounted into the job. The record does not say which input is deserialized or which versions are affected.

Who can reach it

Network vector with user interaction (CVSS UI:R) and no privileges required. In practice this means getting a user or pipeline to process attacker-supplied data with Model Optimizer. The record does not name the exact input path.

What to do

Update NVIDIA Model Optimizer to the release named in NVIDIA bulletin 5914. The record does not give a fixed version. This is a library upgrade in the job environment or container image: rebuild the images and rerun the jobs, no node reboot. Until then, only run Model Optimizer on model and checkpoint files from trusted sources.

References

Related entries

All AI/ML frameworks & serving entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.