CISO Daily Briefing
Cloud Security Alliance Intelligence Report
Executive Summary
AI-assisted intrusion now has a named government victim: the European Commission confirmed an AI-tool campaign, linked by press reports to ARTEX, that disrupted staff systems for over ten days. Attackers are also hitting infrastructure control planes. Puppet CVE-2026-115731 is in KEV with a December 1 federal deadline and can expose the CA key for every managed node, while China-nexus ARMORY CASTLE exploited a vCenter zero-day for two months before the patch. Hunt, not just patch, on both. On the strategic side, the National Cyber Director’s exit creates federal policy continuity risk, and NATO’s reliance on a single AI vendor, OpenAI’s Daybreak, raises concentration questions.
Overnight Research Output
The First Confirmed AI-Tool Attack on an EU Institution: Lessons from the European Commission Incident
CRITICAL URGENCY
Summary: On November 19 the Commission confirmed that a campaign first detected on October 22 was carried out with a publicly available AI tool, disrupting internal HR (Sysper), medical reimbursement (MyPMO) and translation services for more than ten days. Reporting citing Politico sources links it to ARTEX, the autonomous pentest framework that CERT-EU and the FBI warned about in November. Taking down the original repository has not contained it: forks, archives and a commercial “ARTEX PRO” still circulate. Low-priority, internet-facing assets and chains of minor misconfigurations are the primary targets.
Key Sources:
The Record — European Commission confirms cyberattack, says hackers used AI tool
The Register — European Commission scrambles to contain AI-assisted cyberattack
Help Net Security — European Commission confirms AI-generated cyberattack disrupted its systems
FBI FLASH — Threat Overview of ARTEX Autonomous Penetration Testing Framework
Puppet CVE-2026-115731 in KEV: When the Configuration Server Becomes the Fleet Kill Switch
CRITICAL URGENCY
Summary: CVE-2026-115731 lets any client holding a Puppet CA-signed certificate execute code on Puppet Server as the puppet user, which can read the CA signing key. Neodyme showed a stolen key can drive Puppet Enterprise orchestration to run root commands on managed nodes. Where autosigning is enabled, anyone reaching TCP 8140 can enroll and exploit. Companion flaw CVE-2026-115732 (CVSS 9.0) enables node impersonation. Fixes for Puppet Enterprise, Puppet Core and OpenVox shipped November 3; community open-source Puppet back to 6.24.0 has no patch coming.
Key Sources:
CISA — CISA Adds One Known Exploited Vulnerability to Catalog
The Hacker News — Hackers Exploiting Critical Puppet RCE Flaw
Armory Castle’s vCenter Zero-Day: Chained Remote-Support and Directory-Service Exploits Against the Virtualization Control Plane
HIGH URGENCY
Summary: GTIG and Mandiant report that ARMORY CASTLE (formerly UNC5221) exploited CVE-2026-123890 in the VMware Directory Service from early September. The actor entered via BeyondTrust Remote Support CVE-2026-102764, moved across under-monitored Linux appliances, took over vCenter, and copied Active Directory databases directly from guest VM disks on ESXi. The tooling steals the STS SAML signing key, enabling forged admin sessions that bypass MFA and survive password resets until signing material is rotated. Broadcom fixed the flaw in VMSA-2026-0009 on November 19.
Key Sources:
Google Cloud — ARMORY CASTLE Exploits vSphere
Rapid7 — CVE-2026-123890: Critical Authentication Bypass in VMware vCenter
SecurityWeek — Newly Patched VMware vCenter Flaw Exploited as Zero-Day by Chinese APT
Cybersecurity Dive — Chinese espionage group exploits vCenter zero-day
A Vacant Chair at ONCD: US AI-Cyber Policy Continuity After the National Cyber Director’s Exit
HIGH URGENCY
Summary: Cairncross’s tenure ended November 19 after a week of correlated outages across cloud, aviation, banking and national-security networks, none yet tied to a common cause. CyberScoop and CBS report he was fired, and President Trump said he “didn’t think he was good enough,” while Axios sources called it amicable. CSA assesses that binding instruments (EO 14409 deadlines, CISA directives, the Gold Eagle clearinghouse) will hold, but ONCD-coordinated work such as the AI security policy framework and the covered frontier model process faces continuity risk.
Key Sources:
CyberScoop — Sean Cairncross out as national cyber director
NBC News — White House national cyber director is leaving after series of tech outages
The National — Trump didn’t think fired US cyber director was ‘good enough’
Lawfare — AI Cyber Risks Are Testing the Office Built to Coordinate Them
Allied Cyber Defense on a Single Vendor: NATO, ENISA and the Daybreak Concentration Question
HIGH URGENCY
Summary: OpenAI announced on November 17 that NCIA gives NATO’s core institutions Daybreak access, without disclosing models, commercial terms, hosting or continuity arrangements. Access requires case-by-case approval, zero data retention is not default, and models carry removal dates; June’s Anthropic directive showed home governments can also withdraw access. By contrast, the ENISA/CERT-EU joint advisory ran OpenAI and Anthropic models side by side with open-source harnesses, reporting a 95% true-positive rate on 88 owner-assessed findings.
Key Sources:
OpenAI — Strengthening NATO and allied cyber defenses with AI
ENISA & CERT-EU — Joint Advisory on the Use of Frontier AI Models for Code Review
Help Net Security — OpenAI cyber models and NATO
Defence Industry Europe — NATO uses OpenAI models to protect alliance networks
Notable News & Signals
Cisco Expressway CVE-2026-76466 Exploited After PoC Release
Researchers report in-the-wild exploitation of this CVSS 9.8 unauthenticated root RCE in clustered Expressway deployments; upgrade to X15.5.20, as Cisco lists no workaround.
Apache Roller XML-RPC Deserialization RCE Under Attack
Apache warns CVE-2026-129448, an unauthenticated CVSS 9.8 RCE affecting Roller 4.0.0 through 6.1.4, is being exploited; upgrade to the current 6.1.x release.
Russian Espionage Group Pivots Cloud-to-Cloud via Breached Tenants
GTIG reports COLD RELIC impersonating employees of already-breached organizations to phish new targets, then running domain-wide cloud data exports.
Palo Alto Networks to Acquire Applied Compute for $4.5B
The all-stock deal adds open-weight model post-training and inference to Palo Alto’s platform, signaling security vendors moving to build AI, not just guard it.
Topics Already Covered (No New Action Required)
- Meta Muse egress-gate bypass: Covered by “Breaking the Agent’s Egress Gate: Lessons from the Muse Sentinel Bypass” (Nov 21).
- MONAI Label zero-day exploited by AI agents: Amazon Threat Intelligence findings covered Nov 21.
- Signal screenshot blocking vs. OS-level AI agents: Covered by “When the OS Reads the Screen” (Nov 21).
- NGINX2Root / F5 NGINX HTTP/2 RCE: Covered Nov 20.
- Microsoft agent-framework PyPI compromise and Java4Shell (CVE-2026-114938): Covered Nov 19.
- OPTIMJACK / Optimizely client-side supply chain: Covered Nov 18.
- Google GTIG exploit_agent: AI agent writing and launching exploits, covered Nov 18–19.
- GitHub Agentic Autofix accountability: Covered by “Who Owns the Bad Patch?” (Nov 15).
- Fireworks AI → Cursor trojanized Claude Code extension: Covered Nov 16–17.
- ENISA State of Cybersecurity in the Union: Covered Nov 10, 16 and 19.
- Deferred / unverified: SonicWall Gen 6 end-of-support and third-party breaches at JPMorgan, Visa and Mastercard following N0n claims remain on watch. Linux KVM CVE-2026-99312 was dropped because sources conflict on impact, and Bazuka Android malware was dropped for lack of a verifiable source.