NVIDIA GPU Driver EUVD-2026-90140

Requires the attacker to already have unprivileged local access (an authenticated low-privilege account or local code execution) on a Windows or Linux system with the NVIDIA GPU display driver / kernel-mode layer installed and loaded (a GPU must be present and the driver active). Exploitation depends on reaching the kernel-mode error-handling path and forcing the failure condition that causes the improperly initialized resource to be used, so success may be probabilistic rather than guaranteed. Limiting factors: no remote/network vector (AV:L), no unauthenticated access (PR:L), and the flaw is only reachable on hosts running the affected NVIDIA driver - highest concern on shared/multi-tenant GPU hosts (VDI, cloud GPU, vGPU guest driver) where untrusted local users exist.

CVSS 3.1 base score is 7.8 (HIGH) with vector AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, indicating a local attack vector requiring low (but non-zero) privileges and no user interaction, yielding full confidentiality, integrity, and availability impact within the kernel-mode driver's scope. This is a classic local privilege-escalation profile, NOT a remotely exploitable network flaw, so the 'RCE' tag is misleading - the attacker must already have code execution as an unprivileged local user on the machine before this bug is reachable. No EPSS score, CISA KEV listing, or public POC is present in the provided data: exploitation status is therefore 'no public exploit identified at time of analysis.' The CWE-908 (Use of Uninitialized Resource) via an error-handling path typically requires winning a specific failure condition, which can add real-world exploitation friction not fully captured by AC:L. Net assessment: a genuine and patch-worthy local EoP in a very widely deployed driver, but a second-stage/post-compromise issue rather than an internet-facing emergency - prioritize on multi-user, VDI, and cloud-GPU hosts where untrusted local users share GPU hardware.

An unprivileged local user (or malicious process) on a Windows or Linux host running the NVIDIA GPU display driver issues crafted requests to the kernel-mode driver that force an internal operation to fail, driving execution down an error-handling path that acts on a resource that was never properly initialized. By controlling the contents of that uninitialized memory/handle, the attacker corrupts kernel state to escalate to SYSTEM/root, crash the machine, or leak kernel data. No public exploit code is identified at this time, and the local attack vector means the attacker must already be able to run code on the target.