Three Critical Hyper-V CVEs in September 2026 Patch Tuesday Enable Guest-to-Host VM Escape

Microsoft’s September 2026 Patch Tuesday was the largest security release in the company’s history, addressing more than 974 CVEs, and buried in that record volume are three Critical Hyper-V vulnerabilities that deserve attention independent of the overall patch count. CVE-2026-69603, CVE-2026-80083, and CVE-2026-72961 all affect the boundary between guest virtual machines and the Hyper-V host, and two of the three allow a guest to reach code execution on the host itself, the failure mode that turns a single compromised VM into a compromised virtualization host and every co-resident guest on it.

What each vulnerability actually does

• CVE-2026-69603 (CVSS 8.8): a heap-based buffer overflow triggered by a malformed hypercall payload size, allowing arbitrary code execution. Hypercalls are the mechanism a guest uses to request privileged operations from the hypervisor, so a flaw reachable through a malformed hypercall payload is reachable from inside a guest, by design, without any special guest-side privilege beyond running code there at all.
• CVE-2026-80083 (CVSS 8.8): an untrusted pointer dereference reachable from a guest application, also resulting in arbitrary code execution on the host. Specially crafted code running inside a Hyper-V guest can use this to reach code execution on the host, which is the textbook definition of a guest-to-host VM escape.
• CVE-2026-72961 (CVSS 8.2): an elevation-of-privilege flaw involving virtual TPM state data. An attacker with administrative access to a Hyper-V host can supply specially crafted virtual TPM state to a VM in a way that crosses a security boundary and grants Virtual Trust Level 1 (VTL1) privileges, the elevated trust tier used by Windows Virtualization-Based Security features. This one requires existing host admin access rather than being reachable purely from inside a guest, but it still represents a real security boundary failure within the virtualization stack.

Why the guest-to-host pair matters more than the raw CVSS numbers suggest

Most vulnerability triage instinctively ranks flaws by CVSS score, and by that measure these three sit alongside plenty of other Critical items in a record-setting 974-CVE month. What the score alone does not convey is the blast radius specific to virtualization escape flaws. A vulnerability in a single application affects that application. A guest-to-host escape in Hyper-V affects every virtual machine on that host, including ones with no relationship to whatever workload the attacker initially compromised to get code execution inside the vulnerable guest in the first place. If you run any multi-tenant Hyper-V environment, or simply host workloads of meaningfully different trust levels on the same physical hosts, which describes most production virtualization deployments, CVE-2026-69603 and CVE-2026-80083 specifically should be treated with more urgency than their CVSS scores alone might suggest relative to the hundreds of other items in this month’s release.

What to actually do

Apply September 2026’s cumulative security updates to every Hyper-V host in your environment, not just a representative sample. Because the exact cumulative update KB varies by Windows Server and Windows client version, verify the specific update package for each OS version you run against Microsoft’s Security Update Guide directly rather than assuming a single KB number covers your whole fleet. Prioritize hosts running workloads of mixed trust levels, shared hosting environments, hosts running any VM with external or less-trusted network exposure, and hosts running third-party or less-controlled guest workloads, ahead of hosts running only fully-trusted, internally-managed VMs, if you need to sequence patching across a large fleet rather than patching everything simultaneously.

Frequently asked questions

Do these vulnerabilities affect VMware or other hypervisors?
No. CVE-2026-69603, CVE-2026-80083, and CVE-2026-72961 are specific to Microsoft’s Hyper-V hypervisor implementation. They do not affect VMware ESXi, Proxmox VE, or other hypervisor platforms, though guest-to-host escape as a vulnerability class is a risk inherent to virtualization generally, not unique to any one vendor.

Is there a workaround if I cannot patch a Hyper-V host immediately?
Microsoft’s guidance for this release is to apply the security updates, there is no documented workaround for these specific flaws. As a partial mitigation while patching is scheduled, avoid running untrusted or lower-trust guest workloads on hosts that also run sensitive workloads until patched, since that reduces the practical impact of a successful guest-to-host escape even though it does not close the underlying vulnerability.

Does CVE-2026-72961 require a compromised guest to exploit?
No, it works differently from the other two. It requires an attacker who already has administrative access to the Hyper-V host itself, then uses crafted virtual TPM state data to escalate to Virtual Trust Level 1 privileges. It is a privilege escalation within an already-compromised or malicious-admin scenario, not a guest-originated escape like the other two.