CVE-2025-33236: 
NixOS vulnerability analysis and mitigation

CVE-2025-33236 is a code injection vulnerability in the NVIDIA NeMo Framework that allows a low-privileged local attacker to execute arbitrary code by supplying malicious data. It affects all NVIDIA NeMo versions prior to 2.6.1. The vulnerability was disclosed on February 18, 2026, with an initial NVD analysis completed on February 20, 2026. It carries a CVSS v3.1 base score of 7.8 (High), assigned by NVIDIA Corporation (NVIDIA Advisory, NVD).

The vulnerability is classified as CWE-94 (Improper Control of Generation of Code / Code Injection), meaning the framework fails to properly sanitize or restrict attacker-controlled data before it is processed in a code-generation or evaluation context. An attacker with local, low-privileged access can craft malicious input data that, when processed by the NeMo Framework, triggers arbitrary code execution without requiring any user interaction. The attack complexity is low and no special configuration is needed beyond having access to a system running a vulnerable NeMo version. No public proof-of-concept or detailed technical write-up has been published as of the latest available information (NVIDIA Advisory, Cato Networks).

Successful exploitation can result in arbitrary code execution, privilege escalation, information disclosure, and data tampering on the affected host. Because the attack vector is local with low privilege requirements and no user interaction, a compromised low-privileged account or a malicious dataset introduced into an AI/ML pipeline could be sufficient to fully compromise the system running NeMo. The high confidentiality, integrity, and availability impact ratings indicate that all data accessible to the NeMo process — including model weights, training data, and system credentials — could be exposed or corrupted (NVIDIA Advisory, NVD).

There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation as of the latest available information. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities (KEV) catalog. The EPSS score is approximately 0.018%, reflecting a currently low probability of exploitation in the near term. Threat intelligence from Cato Networks CTRL has highlighted this vulnerability alongside related AI/ML framework issues, suggesting growing researcher interest in this attack surface (Cato Networks, NVIDIA Advisory).

NVIDIA has released a patch in NeMo Framework version 2.6.1, which resolves this vulnerability. Users should upgrade to version 2.6.1 or later as the primary remediation step. As interim measures, restrict local access to systems running vulnerable NeMo versions, apply the principle of least privilege to accounts that interact with the framework, and avoid processing untrusted or externally sourced datasets on unpatched systems (NVIDIA Advisory).

Cato Networks CTRL published a threat intelligence report highlighting CVE-2025-33236 alongside related vulnerabilities in AI/ML frameworks such as Meta PyTorch, noting the growing security risk posed by code injection flaws in widely used AI development tools (Cato Networks). Social media activity on Mastodon and other platforms noted the disclosure, though broader community discussion has been limited given the absence of a public exploit.