Platform
python
Component
megatron-lm
Fixed in
0.13.2
CVE-2025-23349 describes a code injection vulnerability discovered in NVIDIA Megatron-LM, a framework for large language model training. Successful exploitation could lead to unauthorized code execution and compromise system integrity. This vulnerability affects all versions of Megatron-LM prior to 0.13.1 and 0.12.3. A patch is available in version 0.13.1.
The vulnerability resides within the tasks/orqa/unsupervised/nq.py component of Megatron-LM. An attacker who can manipulate input to this component can inject malicious code, potentially gaining control of the system running Megatron-LM. This could involve executing arbitrary commands, accessing sensitive data used in model training, modifying training data to influence model behavior, or escalating privileges to access other resources on the system. The impact is particularly severe in environments where Megatron-LM is used for sensitive data processing or critical infrastructure.
CVE-2025-23349 was published on 2025-09-24. Currently, there are no publicly available proof-of-concept exploits. The EPSS score is pending evaluation. It is recommended to monitor security advisories and threat intelligence feeds for any updates regarding active exploitation campaigns.
Exploit Status
EPSS
0.02% (6% percentile)
CISA SSVC
CVSS Vector
The primary mitigation is to upgrade to NVIDIA Megatron-LM version 0.13.1 or later, which contains the fix. If immediate upgrading is not possible, consider isolating instances running vulnerable versions of Megatron-LM to limit the potential blast radius. Review and restrict access to the tasks/orqa/unsupervised/nq.py component. Implement input validation and sanitization to prevent malicious code injection attempts. Monitor system logs for suspicious activity related to Megatron-LM processes.
Actualice NVIDIA Megatron-LM a la versión 0.13.1 o superior. Si no es posible actualizar inmediatamente, considere aplicar las mitigaciones recomendadas por NVIDIA. Revise y valide las entradas del componente tasks/orqa/unsupervised/nq.py para evitar la inyección de código.
Vulnerability analysis and critical alerts directly to your inbox.
CVE-2025-23349 is a code injection vulnerability affecting NVIDIA Megatron-LM versions before 0.13.1 and 0.12.3, allowing potential code execution and data compromise.
You are affected if you are using NVIDIA Megatron-LM versions prior to 0.13.1 or 0.12.3. Check your version and upgrade immediately if vulnerable.
Upgrade to NVIDIA Megatron-LM version 0.13.1 or later to resolve the vulnerability. Consider temporary isolation and input validation as interim measures.
Currently, there are no confirmed reports of active exploitation, but it's crucial to apply the patch promptly to mitigate potential risks.
Refer to the NVIDIA security bulletin for detailed information and updates regarding CVE-2025-23349: [https://nvidia.github.io/megatron-lm/security/advisories/](https://nvidia.github.io/megatron-lm/security/advisories/)
Upload your dependency file and we'll tell you instantly if this and other CVEs hit you.
Upload your requirements.txt file and we'll tell you instantly if you're affected.