VMware Workstation and Fusion CVE-2026-59346 is a critical, CVSS 9.3 integer-overflow flaw in the VMXNET3 virtual network adapter that lets an attacker with local administrative privileges inside a virtual machine execute code on the host, a classic VM escape. Broadcom released fixes on September 5, 2026 and credited @h4urek, @cameudis, and Stan S for the discovery. Because the values are freshly published, NVD has not yet populated the entry, so the 9.3 score here is Broadcom's own rating from the advisory.

Broadcom patched a second flaw in the same release, CVE-2026-59347, a CVSS 8.1 stack-based buffer overflow in HGFS that a local administrator inside a VM can use to run code as the host's VMX process. Both bugs affect VMware Workstation and Fusion versions 25H2 and 26H1, both are fixed in Workstation 26H1u1 and Fusion 26H1u1, and Broadcom says there are no workarounds. Tencent Xuanwu Lab's Yeonghyeon Choi and Tianchu Chen were credited for the HGFS bug.

Both vulnerabilities require the attacker to already hold local administrative privileges inside a guest VM. That precondition sounds high, but it is routinely met: an attacker who phishes a user, exploits weak configuration, or compromises a workload inside a VM can then use these flaws to break out of the guest and reach the host that runs it. Broadcom noted no evidence of exploitation in the wild, but stressed that VMware products are a persistent attack magnet.

What is a VM escape and why does it matter here?

A VM escape is when code running inside a guest virtual machine breaks the isolation boundary and executes on the underlying host. That boundary is the entire security promise of virtualization, so an escape collapses the separation between an untrusted guest and everything else the host runs. CVE-2026-59346 turns local admin in one VM into code execution on the host, which on a shared or lab machine can mean access to other guests and the operator's environment.

How does this connect to the recent vCenter attacks?

It is a reminder that VMware's stack is under sustained pressure. Only last month, attackers actively exploited two vCenter flaws, CVE-2026-59309 and CVE-2026-59310, with the latter suspected to be weaponized by a China-nexus APT; that activity began five days after disclosure and is estimated to have breached 361 unique victim IP addresses across 47 countries, concentrated in Germany, the U.S., Turkey, Iran, and France. The pattern is short disclosure-to-exploitation windows, so treating a 9.3 host-escape as a patch-soon item rather than patch-now is a gamble.

Our read

A VM escape is one of the highest-value bugs an attacker can hold, because it defeats the isolation that virtualization exists to provide. The local-admin precondition is not the safety net it appears to be: real intrusions routinely reach admin inside a workload first, and the escape is the step that turns one compromised VM into control of the host and its neighbors. With VMware's vCenter flaws already exploited weeks earlier within days of disclosure, the defensible posture is to patch Workstation and Fusion to 26H1u1 now, and to verify that a compromised guest genuinely cannot reach the host, rather than assuming the hypervisor boundary holds.

Reporting by The Hacker News; CVSS and fix data per Broadcom's advisory; vCenter exploitation figures per prior reporting. Sources linked above.

Related: What is a VM escape? and What is privilege escalation?.