NVIDIA has released a new software update for its NeMo Framework, addressing multiple high-severity vulnerabilities that could allow attackers to execute arbitrary code, escalate privileges, steal sensitive information, and tamper with data. The flaws impact all platforms and require urgent patching by organizations and developers relying on NeMo for AI and natural language processing (NLP) workflows.
The bulletin details four issues tracked as CVE-2025-23312, CVE-2025-23313, CVE-2025-23314, and CVE-2025-23315, each rated 7.8 (High) on the CVSS scale.
According to NVIDIA, “NVIDIA NeMo Framework for all platforms contains a vulnerability in the retrieval services component, where malicious data created by an attacker could cause a code injection. A successful exploit of this vulnerability might lead to code execution, escalation of privileges, information disclosure, and data tampering.”
Additional flaws were found in the NLP component and the export and deploy component, all of which could be exploited through maliciously crafted input to trigger code injection attacks with similar impact.
If left unpatched, these vulnerabilities could be exploited in environments where NeMo is used for AI model training, deployment, and natural language processing pipelines. Attackers could:
- Execute arbitrary malicious code inside AI workflows.
- Escalate privileges to gain broader system control.
- Access sensitive credentials or training data.
- Tamper with AI datasets or outputs, potentially poisoning models and corrupting decision-making systems.
The potential downstream consequences for organizations using NeMo in production are significant, especially in industries where AI-driven decisions affect finance, healthcare, or security operations.
The vulnerabilities affect all versions of NVIDIA NeMo Framework prior to 2.4.0 across Windows, Linux, and macOS. NVIDIA has released version 2.4.0, which includes fixes for all four CVEs.
Related Posts:
- NVIDIA NeMo Framework: High-Risk Vulnerabilities Allow Remote Code Execution
- NVIDIA Warns of a High-Severity Flaw in NeMo AI Curator (CVE-2025-23307)
- Juniper Security Director Alert: Critical Flaw Allows Unauthenticated Access to Sensitive Resources
Support Our Threat Intelligence
If you find our CVE report and cybersecurity news helpful, consider supporting our work.