Zapscape Flaw in KVM Breaks Nested Virtualization Isolation Allowing L1 Guest Root Code Execution on Linux Host
A security flaw in KVM/x86 within the Linux kernel has highlighted a risk that many infrastructure platforms accept almost without question: nested virtualization exposed to untrusted workloads. The vulnerability, named Zapscape and registered as CVE-2026-64561, can allow an attacker with kernel privileges inside a level-1 virtual machine to execute code on the host with root permissions.
The problem centers on the shadow MMU and the way KVM manages shadow page tables when the L1 guest itself acts as a hypervisor for an L2 guest. In practical terms, the isolation that normally protects the host from activity inside a VM is weakened exactly in the scenario where a provider grants more power to a tenant by allowing nested virtualization inside the tenant’s own VM.
The technical root cause lies in an incorrect ordering when checking a “stale root” condition during guest page-fault handling. This sequence can result in a use-after-free, followed by writes after free — a classic pattern used to turn a memory corruption issue into arbitrary code execution.
Public disclosure includes a proof-of-concept exploit that chains the vulnerability and ultimately creates a file named /Zapscape on the host with root ownership. On Intel platforms the scenario requires the L1 guest to have EPT page walk lengths 4 and 5 exposed. No equivalent condition appears in the published information for AMD processors.
No active exploitation in real-world attacks has been observed so far. Nevertheless, the potential scope justifies rapid action, especially in multi-tenant environments. The NVD lists affected versions starting from Linux 5.9, and fixes are already present in the mainline tree and stable branches, including versions 6.6.148, 6.12.101, 6.18.42, 7.1.6 and 7.2 rc5.
The most direct mitigation is to update to a patched stable kernel or install the vendor package containing the backported fix. At the same time, organizations should inventory hosts that offer nested virtualization to untrusted tenants and treat them as patching priorities. If a server does not require nested virtualization, disabling it immediately reduces the attack surface. Administrators should also verify the effective version through their distribution vendor, because some fixes arrive as backports without obvious changes to the upstream version string.
Related articles
Weekly Roundup: Next.js Leads Critical Vulnerability Reports as Chrome, Tomcat, and PostgreSQL See Major Patches
Security NEXT has published its list of the ten most-read articles for the week of August 23-29, 2026. The top story covered multiple serious vulnerabilities in Next.js that prompted an accelerated release schedule. Other high-interest items included a critical flaw in Zscaler Client Connector, several issues in Apache Tomcat rated critical, and Google’s Chrome 152 update fixing 327 vulnerabilities. Additional reports highlighted a ransomware incident at a systems development firm, flaws in SKYSEA Client View, 28 vulnerabilities fixed in PostgreSQL, and confirmed exploitation of six known flaws in SQL Server and NetScaler. The list also featured a vulnerability in the Android version of the Digital Agency’s My Number Point app and warnings about active exploitation of a WebLogic module flaw.
Multiple Critical Vulnerabilities in Langflow Enable Remote Code Execution
IBM disclosed three severe vulnerabilities in the open-source Langflow project on August 24, 2026. The flaws affect server instances and allow attackers to execute arbitrary code under specific conditions. CVE-2026-19286 is a code injection issue in the Agent-to-Agent (A2A) public endpoint that bypasses existing restrictions when authentication is disabled. CVE-2026-19295 permits authenticated users to evade custom component restrictions via eval injection in flow templates, leading to OS-level command execution. CVE-2026-18729 stems from insufficient input validation and lets low-privileged authenticated users bypass protected paths to run arbitrary code on the server. Patches addressing all three issues are now available.
ServiceNow Discloses Three CVSS 10.0 Vulnerabilities Allowing Unauthenticated Remote Code Execution and SQL Injection
ServiceNow has released security updates addressing four vulnerabilities in its AI platform, including three rated CVSS 10.0. The flaws enable unauthenticated attackers to achieve remote code execution, privilege escalation, and arbitrary SQL execution against core ITSM systems used by large enterprises worldwide. Affected components include the GraphQL Composite Data API, system configuration image upload processor, and dynamic schema ORDER BY handling. ServiceNow states it has patched hosted instances and provided hotfixes for self-hosted customers running Xanadu, Yokohama, Zurich, and Australia branches. This follows a July disclosure of a related sandbox escape tracked as CVE-2026-6875 that showed signs of in-the-wild exploitation. No public exploits or confirmed active attacks have been observed for the new issues yet, but the extremely low attack complexity leaves a narrow remediation window for organizations running exposed instances.
AI Agent Uncovers Unauthenticated Router Config Dump Leading to CVE Filing
An LLM agent tasked only with documenting network topology independently discovered a critical authentication bypass in a home router firmware. The agent performed read-only reconnaissance, extracted the full configuration including base64-encoded admin passwords and WPA2 keys via an unprotected CGI endpoint, and verified the finding by obtaining a valid session cookie. It then produced a complete coordinated disclosure report, classified the issue under CWE-306 with a CVSS 3.1 score of 8.8, and prepared the MITRE CVE submission package. The vulnerability affects LAN-side management interfaces of certain SOHO routers running legacy Boa web servers and remains unpatched due to inaccessible vendor firmware channels. The researcher maintained strict read-only permissions for the agent throughout the process, ensuring no configuration changes occurred. The case demonstrates how autonomous agents can accelerate vulnerability research while staying within defined operational boundaries.