Skip to content
CVEvulnerabilityexploitpatchAI Security

CISA KEV November 9, 2026: four CVEs, November 12 deadline, and an AI infrastructure first

3 min read
Share

CISA added four CVEs to the Known Exploited Vulnerabilities catalog on November 9, 2026, with a Federal Civilian Executive Branch patch deadline of November 12. The batch spans Ivanti Connect Secure, Palo Alto Networks PAN-OS, OpenSSH, and GPUStack. The GPUStack entry is the first time an AI infrastructure component has appeared in the KEV catalog, making this a notable batch beyond its standard severity.

The four CVEs

CVE-2026-102895 (Ivanti Connect Secure): Heap overflow in the SAML parsing component. CVSS 9.1. Exploited in the wild by state-linked actors prior to patch availability. This is Ivanti's seventh CVE to appear in the KEV catalog in 2026.

CVE-2026-116779 (Palo Alto Networks PAN-OS): Privilege escalation via crafted XML in the management interface. CVSS 8.8. Active exploitation observed in targeted campaigns against organizations running internet-exposed PAN-OS management interfaces.

CVE-2026-123158 (OpenSSH, SSHole): Pre-auth remote code execution in the sftp-server component on glibc-based Linux systems. CVSS 9.8. Active exploitation in the wild confirmed. This is the highest-severity CVE in this batch and the first OpenSSH entry in the KEV catalog since 2024.

CVE-2026-123218 (GPUStack): Authentication bypass in the admin API of GPUStack, an open-source GPU cluster manager used to run self-hosted LLM inference. CVSS 8.9. Exploitation requires network access to the admin port. This is the first AI infrastructure component to appear in the KEV catalog.

GPUStack: the AI infrastructure first

GPUStack's appearance in the KEV catalog is a leading indicator. Organizations have been standing up self-hosted GPU inference clusters to run open-weight models, often with the same speed and security shortcuts that characterized early cloud deployments. An authentication bypass in the admin API is a critical vulnerability in that environment: an attacker who gains access to the GPU cluster admin interface can redirect model serving, exfiltrate model weights, or use the cluster's compute for LLMjacking. If you run self-hosted GPU inference, patch GPUStack immediately and audit whether the admin port is network-accessible from untrusted segments.

OpenSSH SSHole: a wake-up for infrastructure teams

CVE-2026-123158 is a pre-auth RCE in OpenSSH's sftp-server on glibc-based Linux, with active exploitation confirmed. The last OpenSSH entry in KEV was in 2024. SFTP exposure is widespread in environments that use it for managed file transfer, secure drop zones, or build artifact delivery. Any glibc-based Linux system running OpenSSH's sftp-server is affected until patched. Priority: treat this as the highest-urgency item in the November 9 KEV batch.

Patch deadlines and prioritization

FCEB agencies must remediate all four CVEs by November 12. For non-federal organizations, all four have confirmed active exploitation, which means the FCEB deadline is a reasonable benchmark for your own patching timeline. Recommended priority order: OpenSSH (pre-auth RCE, highest severity, widest surface), GPUStack (novel AI infrastructure exposure), Ivanti (recurring high-severity, state-actor exploitation), PAN-OS (privilege escalation, management interface access required).

Action items

  • Patch OpenSSH on all glibc-based Linux systems running sftp-server immediately. This is the highest-priority item in the November 9 KEV batch.
  • If you run GPUStack for self-hosted GPU inference, patch immediately and verify that the admin API port is not accessible from untrusted networks. If your GPUStack admin port is internet-exposed, treat that as an active incident.
  • Apply Ivanti Connect Secure and PAN-OS patches in your next maintenance window. All four CVEs have confirmed active exploitation: the November 12 deadline is a reasonable target for all organizations, not just FCEB agencies.

Gigia Tsiklauri is a Security Architect and founder of Infosec.ge. Get in touch if you need help prioritizing these patches or securing AI infrastructure against newly disclosed vulnerabilities.