CVE-2026-61772

NVIDIA Megatron Bridge, a framework in the NVIDIA NeMo ecosystem used for training and fine-tuning large language models, contains a deserialization of untrusted data flaw (CWE-502). Per the CVSS 3.1 vector (AV:L/AC:L/PR:L/UI:N), a local attacker with limited privileges can trigger the flaw by getting the software to deserialize crafted untrusted data, with no user interaction required. A successful exploit may lead to arbitrary code execution, tampering with data, and disclosure of sensitive information, with high impact on confidentiality, integrity, and availability. Potentially affected are environments running Megatron Bridge, such as AI engineering workstations, notebooks, and training clusters, though local access with limited privileges is required to exploit it. There is no known exploitation so far: no public proof-of-concept exists, the issue is not in CISA's KEV catalog, and EPSS assigns only a 0.2% probability of exploitation in the next 30 days.

What to do: Check NVIDIA's PSIRT security advisory for the exact affected version range and upgrade Megatron Bridge to the fixed release it specifies, since fixed versions are not given in the source data. Until patching, avoid deserializing untrusted inputs such as checkpoints, pickled artifacts, or serialized job data from untrusted sources, and restrict local access to systems running Megatron Bridge. Given the high severity rating but absence of known exploitation and low EPSS, treat this as a standard priority patch rather than an emergency, and monitor NVIDIA advisories for updates.