Azure OpenAI just had a CVSS 9.9 vulnerability. Any authorized user could have escalated to your entire network.

Microsoft just patched a server-side request forgery in Azure OpenAI.

CVSS 9.9. Critical.

The flaw: an authorized user with low privileges could route requests through Azure OpenAI to reach internal metadata endpoints, cloud credentials, and restricted network resources. The AI service became a proxy into your infrastructure.

No admin access needed. Just a valid API key and a crafted request.

Microsoft fixed it server-side. No customer action required. That should terrify you.

Because it means your AI platform had a privilege escalation path you never knew about. You didn't patch it. You didn't detect it. Microsoft just flipped a switch and it was gone.

Ask yourself: how many other silent failure modes exist in the AI services your developers deployed without security review?

This is the third major AI platform vulnerability in 30 days. JADEPUFFER ran autonomous ransomware through Langflow. Vercel got extorted through a shadow AI browser extension. Now Azure OpenAI.

The attack surface isn't your models. It's the infrastructure around them. The API gateways. The configuration stores. The IAM bindings you never audited.

Your CISO needs a full inventory of every AI service touching production data. Not next quarter. This week.

Audit every Azure OpenAI deployment. Review IAM roles. Restrict outbound requests. And start treating AI infrastructure as Tier 1 attack surface — because attackers already do.