📧 info@ciso.sa | 📱 +966550939344 | Riyadh, Kingdom of Saudi Arabia
🔧 Scheduled Maintenance — Saturday 2:00-4:00 AM AST. Some features may be temporarily unavailable.    ●   
💎
Pro Plan 50% Off Unlock all AI features, unlimited reports, and priority support. Upgrade
Search Center
ESC to close
Global data_breach Software/SaaS CRITICAL 1h Global vulnerability Technology / Web Services HIGH 3h Global vulnerability Consumer Electronics / Technology CRITICAL 5h Global ransomware Enterprise / All Sectors CRITICAL 5h Global data_breach Government CRITICAL 7h Global malware Multiple sectors / General public HIGH 7h Global vulnerability Technology and Software Development CRITICAL 8h Global malware,vulnerability,apt Technology, Cloud Services, Consumer Electronics HIGH 8h Global malware Web Hosting and Content Management HIGH 8h Global vulnerability Information Technology and Network Infrastructure CRITICAL 9h Global data_breach Software/SaaS CRITICAL 1h Global vulnerability Technology / Web Services HIGH 3h Global vulnerability Consumer Electronics / Technology CRITICAL 5h Global ransomware Enterprise / All Sectors CRITICAL 5h Global data_breach Government CRITICAL 7h Global malware Multiple sectors / General public HIGH 7h Global vulnerability Technology and Software Development CRITICAL 8h Global malware,vulnerability,apt Technology, Cloud Services, Consumer Electronics HIGH 8h Global malware Web Hosting and Content Management HIGH 8h Global vulnerability Information Technology and Network Infrastructure CRITICAL 9h Global data_breach Software/SaaS CRITICAL 1h Global vulnerability Technology / Web Services HIGH 3h Global vulnerability Consumer Electronics / Technology CRITICAL 5h Global ransomware Enterprise / All Sectors CRITICAL 5h Global data_breach Government CRITICAL 7h Global malware Multiple sectors / General public HIGH 7h Global vulnerability Technology and Software Development CRITICAL 8h Global malware,vulnerability,apt Technology, Cloud Services, Consumer Electronics HIGH 8h Global malware Web Hosting and Content Management HIGH 8h Global vulnerability Information Technology and Network Infrastructure CRITICAL 9h
Vulnerabilities

CVE-2025-12805

High
A flaw was found in Red Hat OpenShift AI (RHOAI) llama-stack-operator. This vulnerability allows unauthorized access to Llama Stack services deployed in other namespaces via direct network requests, b
CWE-653 — Weakness Type
Published: Mar 26, 2026  ·  Modified: Apr 2, 2026  ·  Source: NVD
CVSS v3
8.1
🔗 NVD Official
📄 Description (English)

A flaw was found in Red Hat OpenShift AI (RHOAI) llama-stack-operator. This vulnerability allows unauthorized access to Llama Stack services deployed in other namespaces via direct network requests, because no NetworkPolicy restricts access to the llama-stack service endpoint. As a result, a user in one namespace can access another user’s Llama Stack instance and potentially view or manipulate sensitive data.

🤖 AI Executive Summary

CVE-2025-12805 is a critical namespace isolation bypass in Red Hat OpenShift AI's llama-stack-operator that allows unauthorized cross-namespace access to Llama Stack services. Without NetworkPolicy enforcement, attackers can directly access services in other namespaces, potentially exposing sensitive AI model data and enabling unauthorized manipulation. This vulnerability poses significant risk to multi-tenant Kubernetes environments commonly deployed in Saudi organizations.

📄 Description (Arabic)

🤖 AI Intelligence Analysis Analyzed: Apr 25, 2026 20:21
🇸🇦 Saudi Arabia Impact Assessment
This vulnerability directly impacts Saudi organizations using Red Hat OpenShift AI for AI/ML workloads, particularly in: (1) Banking sector (SAMA-regulated institutions) processing sensitive financial AI models; (2) Government agencies (NCA oversight) running multi-tenant AI platforms; (3) Healthcare organizations managing patient data through AI services; (4) Energy sector (ARAMCO, utilities) deploying predictive analytics; (5) Telecom providers (STC, Mobily) using AI for network optimization. Multi-tenant Kubernetes deployments in Saudi data centers are at highest risk due to namespace isolation assumptions being violated.
🏢 Affected Saudi Sectors
Banking and Financial Services Government and Public Administration Healthcare and Medical Services Energy and Utilities Telecommunications Education and Research Manufacturing and Industrial
⚖️ Saudi Risk Score (AI)
8.4
/ 10.0
🔧 Remediation Steps (English)
IMMEDIATE ACTIONS:
1. Audit all Red Hat OpenShift AI deployments to identify llama-stack-operator instances and document namespace configurations
2. Implement NetworkPolicy restrictions immediately to deny cross-namespace traffic to llama-stack services
3. Review access logs for unauthorized cross-namespace service access attempts
4. Isolate affected Llama Stack instances if multi-tenant sharing is detected

COMPENSATING CONTROLS (until patch available):
5. Deploy Kubernetes NetworkPolicy with default-deny ingress rules, explicitly allowing only same-namespace traffic to llama-stack endpoints
6. Implement service mesh (Istio/Linkerd) with strict mTLS and authorization policies enforcing namespace boundaries
7. Use RBAC to restrict service account permissions across namespaces
8. Enable Kubernetes audit logging for all cross-namespace API calls

DETECTION RULES:
9. Monitor for DNS queries to llama-stack services from different namespaces
10. Alert on successful TCP connections to llama-stack service ports from external namespaces
11. Track API calls attempting to access llama-stack resources across namespace boundaries
12. Monitor for unusual data exfiltration patterns from AI model endpoints

PATCHING:
13. Subscribe to Red Hat security advisories for llama-stack-operator patches
14. Prepare change management process for immediate deployment once patch is released
🔧 خطوات المعالجة (العربية)
الإجراءات الفورية:
1. تدقيق جميع نشرات Red Hat OpenShift AI لتحديد مثيلات llama-stack-operator وتوثيق تكوينات مساحة الأسماء
2. تنفيذ قيود سياسة الشبكة فوراً لرفض حركة المرور عبر مساحات الأسماء إلى خدمات llama-stack
3. مراجعة سجلات الوصول لمحاولات الوصول غير المصرح به عبر مساحات الأسماء
4. عزل مثيلات Llama Stack المتأثرة إذا تم اكتشاف مشاركة متعددة المستأجرين

الضوابط التعويضية (حتى توفر التصحيح):
5. نشر سياسة شبكة Kubernetes مع قواعد الرفض الافتراضية للدخول، مما يسمح صراحة فقط بحركة المرور من نفس مساحة الأسماء إلى نقاط نهاية llama-stack
6. تنفيذ شبكة الخدمات (Istio/Linkerd) مع mTLS صارم وسياسات التفويض التي تفرض حدود مساحة الأسماء
7. استخدام RBAC لتقييد أذونات حساب الخدمة عبر مساحات الأسماء
8. تفعيل تسجيل تدقيق Kubernetes لجميع استدعاءات API عبر مساحات الأسماء

قواعد الكشف:
9. مراقبة استعلامات DNS إلى خدمات llama-stack من مساحات أسماء مختلفة
10. تنبيه الاتصالات الناجحة بـ TCP إلى منافذ خدمة llama-stack من مساحات أسماء خارجية
11. تتبع استدعاءات API التي تحاول الوصول إلى موارد llama-stack عبر حدود مساحة الأسماء
12. مراقبة أنماط تسرب البيانات غير العادية من نقاط نهاية نموذج الذكاء الاصطناعي

التصحيح:
13. الاشتراك في تنبيهات أمان Red Hat لتصحيحات llama-stack-operator
14. تحضير عملية إدارة التغيير للنشر الفوري بمجرد إصدار التصحيح
📋 Regulatory Compliance Mapping
🟢 NCA ECC 2024
ECC 2024 A.8.1 - Asset Management (Kubernetes cluster inventory) ECC 2024 A.13.1 - Network Security (Network segmentation and access control) ECC 2024 A.13.2 - Network Architecture (Namespace isolation enforcement) ECC 2024 A.14.2 - System Monitoring (Detection of unauthorized access attempts)
🔵 SAMA CSF
SAMA CSF ID.AM-2 - Asset Management (Identify all AI/ML systems) SAMA CSF PR.AC-3 - Access Control (Enforce least privilege across namespaces) SAMA CSF PR.DS-2 - Data Security (Protect sensitive AI model data) SAMA CSF DE.CM-1 - Detection and Analysis (Monitor for cross-namespace access)
🟡 ISO 27001:2022
ISO 27001:2022 A.8.1 - Asset Management ISO 27001:2022 A.8.3 - Acceptable use of assets ISO 27001:2022 A.9.1 - Access control policy ISO 27001:2022 A.13.1 - Network security ISO 27001:2022 A.13.2 - Network architecture
🟣 PCI DSS v4.0.1
PCI DSS 1.2 - Network segmentation (if processing payment data through AI) PCI DSS 7.1 - Limit access to cardholder data by business need-to-know
📊 CVSS Score
8.1
/ 10.0 — High
📊 CVSS Vector
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N
Attack VectorN — None / Network
Attack ComplexityL — Low / Local
Privileges RequiredL — Low / Local
User InteractionN — None / Network
ScopeU — Unchanged
ConfidentialityH — High
IntegrityH — High
AvailabilityN — None / Network
📋 Quick Facts
Severity High
CVSS Score8.1
CWECWE-653
Exploit No
Patch ✗ No
Published 2026-03-26
Source Feed nvd
Views 4
🇸🇦 Saudi Risk Score
8.4
/ 10.0 — Saudi Risk
Priority: CRITICAL
🏷️ Tags
CWE-653
Share this CVE

💬 Comments

0
Loading comments
📣 Found this valuable?
Share it with your cybersecurity network
in LinkedIn 𝕏 X / Twitter 💬 WhatsApp ✈ Telegram
🍪 Privacy Preferences
CISO Consulting — Compliant with Saudi Personal Data Protection Law (PDPL)
We use cookies and similar technologies to provide the best experience on our platform. You can choose which types you accept.
🔒
Essential Always On
Required for the website to function properly. Cannot be disabled.
📋 Sessions, CSRF tokens, authentication, language preferences
📊
Analytics
Help us understand how visitors use the site and improve performance.
📋 Page views, session duration, traffic sources, performance metrics
⚙️
Functional
Enable enhanced features like content personalization and preferences.
📋 Dark/light theme, font size, custom dashboards, saved filters
📣
Marketing
Used to deliver content and ads relevant to your interests.
📋 Campaign tracking, retargeting, social media analytics
Privacy Policy →
CISO AI Assistant
Ask anything · Documents · Support
🔐

Introduce Yourself

Enter your details to access the full assistant

Your info is private and never shared
💬
CyberAssist
Online · responds in seconds
5 / 5
🔐 Verify Your Identity

Enter your email to receive a verification code before submitting a support request.

Enter to send · / for commands 0 / 2000
CISO AI · Powered by Anthropic Claude
✦ Quick Survey Help Us Improve CISO Consulting Your feedback shapes the future of our platform — takes less than 2 minutes.
⚠ Please answer this question to continue

How would you rate your overall experience with our platform?

Rate from 1 (poor) to 5 (excellent)

🎉
Thank you!
Your response has been recorded.