CHAIN-147 — OpenAI deployment with no content filter + jailbreak leakage

Summary

Severity: Medium · Likelihood: High · Logic: ALL

Why this chain matters

An Azure OpenAI deployment has content filters disabled (or set to lowest) AND is publicly reachable. Attackers extract system prompts, harvest RAG context, and use the endpoint as a free LLM relay — all invoices paid by the victim.

Component rules

This chain fires when its trigger conditions are met by the following rules. Click any rule to see its detection logic and compliance mappings.

Rule ID	Role
zt_ai_006	Trigger
zt_ai_001	Trigger

Attack walkthrough

Step 1 — Send jailbreak prompt to extract the system prompt.

Actor: Attacker
MITRE ATT&CK: T1557.003
Enabled by: zt_ai_006

Attacker gain: Internal prompt + any RAG data embedded in context.

Step 2 — Relay arbitrary prompts through the endpoint; bill paid by victim.

Actor: Attacker
MITRE ATT&CK: T1496
Enabled by: zt_ai_001

Attacker gain: Free LLM access + potential reputation harm from outputs.

Blast radius

Initial access	Public LLM endpoint.
Max privilege	System prompt + RAG leak; compute abuse.
Data at risk	System prompts, RAG context corpus
Services at risk	OpenAI deployment

How the logic works

The chain fires only when every rule above has at least one finding in the current scan. Missing any one rule breaks the chain — so remediating any single step disrupts the attack path.