Database/AI/ML frameworks & serving
NVIDIA Megatron-LM: unsafe deserialization when loading checkpoints allows code execution
Impact
Megatron-LM deserializes untrusted data when it loads a checkpoint, so a crafted checkpoint can run code as the training process. That process usually holds the job's GPUs, its dataset and storage credentials, and often a shared filesystem mount, so one poisoned checkpoint can tamper with or leak other teams' artifacts on a shared training cluster. NVIDIA rates the impact as code execution, information disclosure and data tampering.
Who can reach it
Local, low privilege: anyone who can place or swap a checkpoint that a Megatron-LM job will load, for example on shared storage, in a model hub mirror or in a hand-off between teams. No user interaction beyond the normal load.
What to do
Upgrade Megatron-LM to the fixed release named in NVIDIA bulletin 5926 and restart the training jobs that use it. The record does not give a fixed version. Until you upgrade, load checkpoints only from trusted, write-controlled locations.
References
Related entries
- NVIDIA Megatron-LM: malicious checkpoint files execute code when loadedCVE-2026-65110 · NVIDIA Megatron-LM (checkpoint loading)High
- NVIDIA Isaac Lab: eval injection allows code execution from attacker-influenced inputCVE-2026-65141 · NVIDIA Isaac LabHigh
- NVIDIA Model-Optimizer: deserialization of untrusted data leads to code executionCVE-2026-65142 · NVIDIA Model-Optimizer (model/checkpoint deserialization)High
- NVIDIA NeMo: a crafted model_config.yaml injects unsafe parameters into dataset loadingCVE-2026-65178 · NVIDIA NeMo (dataset-loading workflow, model_config.yaml parameter injection)High
- PyTorch (flatbuffer model parsing, torch::load / parse_and_initialize_mobile_module): MALICIOUS MODEL FILE TO MEMORYNCVD-2025-019-pytorch-flatbuffer-model-parsing · PyTorch (flatbuffer model parsing, torch::load / parse_and_initialize_mobile_module)High
- LangChain (Web Research Retriever): SSRFCVE-2024-3095 · LangChain (Web Research Retriever)High
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.