NVIDIA NeMo’s dataset‑loading workflow processes user‑supplied model_config.yaml files. A maliciously crafted YAML file can inject unsafe parameters that, once deserialized, allow an attacker to execute arbitrary code, tamper with data, cause a denial of service, or disclose confidential information. The vulnerability is a classic deserialization flaw described by CWE‑502.

This issue affects NVIDIA NeMo Speech. Because the advisory does not specify particular release numbers, all installed versions released before the fix should be considered vulnerable until the vendor provides an update.

The CVSS score of 7.8 indicates a high severity. The EPSS score is not available, and the vulnerability is not listed in the CISA KEV catalog. The likely attack vector is the supply or placement of a malicious model_config.yaml that is processed by the dataset loader, which usually requires either local file system access or influence over a trusted deployment. When the YAML is parsed, unsafe tags can trigger code execution and other impacts.