Microsoft has published CVE-2026-85889, a critical vulnerability in Azure AI Foundry (formerly Azure AI Studio) with a CVSS score of 10.0 — the maximum possible severity rating.
What Is Azure AI Foundry?
Azure AI Foundry is Microsoft's unified platform for building, deploying, and managing AI applications. It provides tools for model training, fine-tuning, deployment, monitoring, and governance. Organizations use it to develop custom AI solutions on Azure infrastructure.
The Vulnerability
CVE-2026-85889 allows a remote, unauthenticated attacker to execute arbitrary code on the Azure AI Foundry service. The vulnerability exists in the API gateway component that handles requests to deployed AI models.
Impact includes:
- Remote code execution on the AI Foundry service
- Model exfiltration — attackers can steal deployed AI models
- Training data exposure — sensitive training data used to fine-tune models can be accessed
- Lateral movement — access to the Azure environment hosting the AI services
CVSS 10.0 Breakdown
A CVSS 10.0 score means the vulnerability has maximum impact across all dimensions:
- Attack Vector: Network — exploitable remotely
- Attack Complexity: Low — no special conditions required
- Privileges Required: None — no authentication needed
- User Interaction: None — no user action required
- Scope: Changed — impact extends beyond the vulnerable component
- Confidentiality Impact: High — sensitive data exposed
- Integrity Impact: High — code execution allows modification
- Availability Impact: High — service can be disrupted
Microsoft's Response
Microsoft has released a patch for CVE-2026-85889 and urges all Azure AI Foundry customers to apply it immediately. The vulnerability affects all Azure AI Foundry deployments regardless of region or configuration.
Broader Context
This vulnerability highlights the security challenges of AI platforms. AI services handle highly sensitive assets — proprietary models, training data, and the AI applications built on top of them. A compromised AI platform can have far-reaching consequences including intellectual property theft and exposure of sensitive data used in model training.
Recommendations
- Apply the Microsoft patch for CVE-2026-85889 immediately
- Review Azure AI Foundry deployment configurations for additional hardening
- Audit access logs for signs of exploitation
- Review AI model and training data for potential exposure