CVE-2026-65141 Overview
NVIDIA Isaac Lab contains an eval injection vulnerability that allows a local attacker to execute arbitrary code through unsafe evaluation of attacker-influenced input. Successful exploitation can lead to code execution, information disclosure, data tampering, and denial of service on the host running Isaac Lab. The flaw is categorized under CWE-95 (Improper Neutralization of Directives in Dynamically Evaluated Code). Exploitation requires local access and user interaction, but no privileges are needed to launch the attack.
Critical Impact
Local eval injection in NVIDIA Isaac Lab can give attackers arbitrary code execution with full compromise of confidentiality, integrity, and availability on the affected host.
Affected Products
- NVIDIA Isaac Lab (robotics simulation and learning framework)
- Refer to the NVIDIA product security advisory for specific affected versions
- See the NVD entry for CVE-2026-65141 for the authoritative product list
Discovery Timeline
- 2026-10-13 - CVE CVE-2026-65141 published to NVD
- 2026-10-13 - Last updated in NVD database
Technical Details for CVE-2026-65141
Vulnerability Analysis
The vulnerability is an eval injection flaw in NVIDIA Isaac Lab, a GPU-accelerated framework used for robot learning and simulation. The affected code path passes attacker-controlled input into a dynamic code evaluation routine without proper neutralization. When the evaluation runs, the injected expression executes within the Python runtime context of Isaac Lab. The result is arbitrary code execution with the same privileges as the Isaac Lab process. The attack vector is local and requires user interaction, such as loading a crafted configuration, script, experiment file, or policy artifact.
Root Cause
The root cause is improper neutralization of directives in dynamically evaluated code, classified under CWE-95. Isaac Lab accepts input values that are later consumed by a dynamic evaluation construct, such as eval(), exec(), or an equivalent interpreter call. Because the input is not restricted to a safe grammar or validated against an allow-list, any Python expression supplied by an attacker executes verbatim. Technical details for the vulnerable code path are referenced in the NVIDIA product security repository.
Attack Vector
An attacker crafts a malicious artifact (configuration file, experiment definition, or similar input) and convinces a legitimate user to load it in Isaac Lab. Upon loading, the embedded expression is evaluated and runs arbitrary commands on the host. The attack requires local access and user interaction, and no authentication is required. Exploitation yields code execution, information disclosure, data tampering, or denial of service, as described in the CVE record for CVE-2026-65141.
Stop Remote Code Execution Before It Runs
Singularity™ Endpoint detects and blocks exploit attempts at machine speed, including zero-days and novel attack techniques.
Stop Exploit AttemptsDetection Methods for CVE-2026-65141
Indicators of Compromise
- Unexpected child processes spawned by the Python interpreter running Isaac Lab (for example, shells, curl, wget, or compilers).
- Isaac Lab configuration, experiment, or policy files sourced from untrusted locations or containing suspicious Python expressions.
- Outbound network connections initiated by the Isaac Lab process to unknown hosts shortly after a user loads a new artifact.
- Modifications to files outside the Isaac Lab working directory immediately following artifact loading.
Detection Strategies
- Hunt for process-tree anomalies where Isaac Lab (python, isaaclab) spawns command interpreters or networking tools.
- Inspect configuration and experiment files for calls to eval, exec, __import__, compile, or os.system in fields that should contain data only.
- Correlate file read events on Isaac Lab inputs with subsequent process creation and network activity on the same host.
Monitoring Recommendations
- Enable endpoint telemetry on hosts running Isaac Lab to capture process lineage, command-line arguments, and script content.
- Alert on Python processes that write to persistence locations such as ~/.bashrc, cron, or systemd user units.
- Monitor egress from GPU workstations and simulation servers for connections to non-corporate destinations.
How to Mitigate CVE-2026-65141
Immediate Actions Required
- Apply the patched Isaac Lab release referenced in the NVIDIA product security advisory.
- Treat all Isaac Lab configuration files, experiment definitions, and policy artifacts from external sources as untrusted until reviewed.
- Restrict Isaac Lab execution to dedicated, least-privilege user accounts without administrative rights.
- Audit shared storage and repositories for artifacts added or modified by unknown users.
Patch Information
NVIDIA has published this vulnerability through its product security program. Review the advisory at the NVIDIA product-security repository entry 5902 for the fixed version and upgrade instructions. Install the patched build on every host that uses Isaac Lab, including developer workstations, training nodes, and simulation servers.
Workarounds
- Only load Isaac Lab artifacts from trusted, version-controlled sources with code review enforced.
- Run Isaac Lab inside an isolated container or virtual machine with no access to sensitive data or credentials.
- Remove or restrict write access to shared configuration directories so unprivileged users cannot plant malicious artifacts.
- Enforce code signing or checksum validation on experiment and policy files before they are loaded.
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
