CVE-2025-23266

A flaw was found in the NVIDIA Container Toolkit. This vulnerability allows execution of arbitrary code with elevated permissions via improperly secured container initialization hooks. This can potentially lead to privilege escalation, data tampering, information disclosure, and denial of service.

RHEL AI is not affected because it uses CDI mode with NVIDIA Container Toolkit versions later than 1.17.5, which are not vulnerable, and does not rely on the enable-cuda-compat hook, thereby eliminating the attack vector entirely.

This vulnerability is marked as Important rather than Critical primarily due to its attack prerequisites and execution context, despite its high CVSS score. While the vulnerability allows privilege escalation via the enable-cuda-compat hook in the NVIDIA Container Toolkit, it is not exploitable remotely—it requires local access and low privileges within a containerized environment. This significantly reduces its initial attack surface compared to a truly critical vulnerability that could be exploited over the network without authentication.

The CVSS score(s) listed for this vulnerability do not reflect the associated product's status, and are included for informational purposes to better understand the severity of this vulnerability.

No mitigation is currently available that meets Red Hat Product Security’s standards for usability, deployment, applicability, or stability.