NVIDIA NVAPI CVE-2026-47573

Requires the NVIDIA GPU driver package (which bundles NVAPI) installed on a Windows host - i.e. GeForce, RTX/Quadro/NVS, or Tesla driver-equipped systems. The attacker must already have LOCAL access with at least low-level user privileges (PR:L) and the ability to invoke NVAPI entry points; no user interaction (UI:N) and no special non-default configuration is called out beyond a standard driver install. Limiting factors: local-only attack vector (no remote/network exploitation), a prerequisite foothold as a standard user, and scope-unchanged impact confined to the host. Realistic exposure is highest on shared/multi-user and GPU-accelerated VDI or cloud instances where untrusted users run code on the same machine.

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, meaning exploitation requires LOCAL access and low-but-nonzero privileges (an authenticated user already on the machine), not remote reach. The full C:H/I:H/A:H impact triad is what drives the 7.8 - a single out-of-bounds write (CWE-787) in NVAPI can lead to code execution, privilege escalation, DoS, info disclosure, and tampering. No EPSS score, no CISA KEV listing, and no public POC are present in the provided data, so there is no evidence of active exploitation or automated weaponization. The 'RCE' tag overstates reach: with AV:L this is local code execution / privilege escalation, not remote code execution. Net assessment: a genuine local privilege-escalation concern worth patching on multi-user, VDI, and cloud-GPU hosts, but NOT an internet-facing emergency - the local-only vector and absence of exploitation signals place it below network-reachable pre-auth flaws in triage priority.

A low-privileged user (or malware already running in a standard user context) on a Windows workstation or GPU-backed VDI/cloud instance calls a crafted NVAPI entry point that triggers an out-of-bounds write in the NVIDIA driver's user-mode API surface. By controlling the overwritten memory, the attacker corrupts state to execute code or escalate privileges, breaking out of the low-privilege sandbox. No public exploit code is identified at time of analysis, and the attack requires prior local code execution.