Windows Hyper-V EUVD-2026-35686

Requires Hyper-V to be installed and enabled on the target Windows 10/11 or Windows Server build, plus a local low-privileged authenticated session on that host (PR:L, AV:L, UI:N) - there is no remote network vector and no user interaction needed once the attacker has a local account. Exploitation is limited by the need for that local foothold, by Hyper-V not being installed on most default client Windows installations, and by SSVC indicating the issue is not automatable; the description does not state whether the 'local' context is the host OS or a Hyper-V guest, so guest-to-host escape cannot be confirmed from the data provided.

Signals are mixed and lean toward 'patch on normal cadence, not emergency': CVSS 7.8 with AV:L/AC:L/PR:L/UI:N reflects easy local exploitation by any authenticated low-privilege user with high impact on C/I/A, and SSVC marks Technical Impact as 'total' - consistent with a hypervisor compromise. However, SSVC Exploitation is 'none' and Automatable is 'no', EPSS is only 0.06% (19th percentile), and there is no KEV listing and no public exploit identified at time of analysis. The 'execute code locally' wording plus AV:L means this is not a remote or guest-to-host escape claim from the description alone; whether the low-privileged context is the host or a Hyper-V guest is NOT explicitly stated in the description and should be verified against the MSRC advisory before triaging as a VM escape. Tag 'Information Disclosure' alongside 'Buffer Overflow' is somewhat inconsistent with code-execution wording - likely reflecting the underlying CWE-125 read primitive - and should not override the description's RCE characterization.

An attacker who already has a low-privileged interactive or service account on a Windows host with Hyper-V enabled triggers the out-of-bounds read in a Hyper-V interface reachable from that context, corrupting hypervisor state to execute attacker-controlled code with elevated privileges. No public exploit identified at time of analysis, and SSVC marks the issue as not automatable, so realistic use is likely chained after an initial foothold (phishing, stolen creds, or another bug) rather than as a standalone mass-exploitation primitive.