TITAN AI · CONDUIT Agent
CONDUIT TICKET REPORT
INCIDENT
Azure
Critical
Closed
Opened by TITAN AI · CONDUIT · Scanner found the finding, CONDUIT forwarded this incident via CONDUIT generic-API layer, assigned security_engineering, populated every ticket field, TITAN FORGE applied the fix, TITAN SCOUT rescanned, and CONDUIT auto-closed the ticket with a Successful close_code.
Finding Summary
- Short description
- [TITAN] Critical — data_leak on openai-banking-copilot
- Severity
- Critical
- Priority
- 1 - Critical
- Resource
- openai-banking-copilot
- Resource type
- Microsoft.CognitiveServices/accounts
- Cloud
- Azure
- Subscription / Account
- 4f29d094-1079-44c9-acb0-4d73a7a2dd34
- Resource group / Project
- rg-banking-prod
The Security Finding
Banking-copilot OpenAI endpoint received 38 prompts containing raw customer SSNs, account numbers, and card PANs in the last 24h. GLBA §501(b) + PCI DSS 3.4 + SOX ITGC violation — card data + PII leaked into third-party LLM prompt logs. AI GUARD blocked exfiltration and rotated prompts.
ITIL Change Management Fields
Justification
Critical: Banking-copilot OpenAI endpoint received 38 prompts containing raw customer SSNs, account numbers, and card PANs in the last 24h. GLBA §501(b) + PCI DSS 3.4 + SOX ITGC violation — card data + PII leaked into third-party LLM prompt logs. AI GUARD blocked exfiltration and rotated prompts.
Implementation Plan
1. Pre-change snapshot captured by TITAN (auto-rollback available).
2. Execute fix command:
az cognitiveservices account network-rule add --name openai-banking-copilot --resource-group rg-banking-ai --ip-address 10.0.0.0/8 && titan-ai-guard apply-policy --policy banking-strict-pii
3. TITAN FORGE verifies the fix was applied.
4. Post-change rescan by TITAN SCOUT — finding must no longer appear.
5. Close ticket with Successful close_code.
Risk & Impact Analysis
Risk level: HIGH business risk — active exposure; fix required immediately.
Business impact if unremediated: Potential data exfil, privilege escalation, or compliance breach.
Scope: single resource (openai-banking-copilot).
Blast radius: change is idempotent; pre-change snapshot captured by TITAN; auto-rollback available if rescan fails.
Finding detail: Banking-copilot OpenAI endpoint received 38 prompts containing raw customer SSNs, account numbers, and card PANs in the last 24h. GLBA §501(b) + PCI DSS 3.4 + SOX ITGC violation — card data + PII leak
Backout / Rollback Plan
1. TITAN auto-captured snapshot of openai-banking-copilot before change (baseline: titan-banking-demo-20260422T201922Z).
2. If post-change rescan still shows the finding OR a new issue appears within 15 min:
a. TITAN FORGE fires rollback automatically using stored snapshot.
b. Incident reopens and escalates to on-call.
3. Manual rollback command path (human override) is documented in close notes.
Test Plan
1. TITAN SCOUT rescans openai-banking-copilot immediately after FORGE applies the change.
2. PASS criteria: the specific finding no longer appears in SCOUT results.
3. PASS criteria: no new CRITICAL or HIGH findings introduced by the change.
4. Automated compliance check: HIPAA/PCI/SOC2 controls re-evaluated.
5. If any check fails, backout plan fires automatically.
Recommended Fix
az cognitiveservices account network-rule add --name openai-banking-copilot --resource-group rg-banking-ai --ip-address 10.0.0.0/8 && titan-ai-guard apply-policy --policy banking-strict-pii
Compliance Mapping
SOC 2 CC7.1HIPAA §164.312(b)
Routing Metadata
- Assignment group
- security_engineering
- Change type
- Emergency
- Approval required
- Yes
- Planned start
- —
- Planned end
- —
- Scan ID
- titan-banking-demo-20260422T201922Z
- Generated at
- 2026-04-22T18:49:22+00:00
- Opened
- 2026-04-22 18:49:22
- Closed
- 2026-04-22 19:03:22
- Close code
- Successful
Attached Security Ticket
SERVICENOW · INCIDENT
SEC-8369 · [TITAN] Critical — data_leak on openai-banking-copilot
Priority: Critical
INCIDENT
Azure
Ticket Description
Banking-copilot OpenAI endpoint received 38 prompts containing raw customer SSNs, account numbers, and card PANs in the last 24h. GLBA §501(b) + PCI DSS 3.4 + SOX ITGC violation — card data + PII leaked into third-party LLM prompt logs. AI GUARD blocked exfiltration and rotated prompts.
Recommended Fix: az cognitiveservices account network-rule add --name openai-banking-copilot --resource-group rg-banking-ai --ip-address 10.0.0.0/8 && titan-ai-guard apply-policy --policy banking-strict-pii
AI Close Notes
TITAN CONDUIT orchestrated end-to-end: SCOUT detected the critical data_leak incident, FORGE applied the consent-gated fix automatically (incident class), SCOUT rescan confirmed the finding cleared, and CONDUIT closed this ticket with a Successful close_code. Pre-change snapshot retained for 30 days for rollback.