[ ABORT TO HUD ]
SEQ. 1
SEQ. 2
SEQ. 3

Agentic Data Loss Prevention (DLP) & Prompt Shields

🛡️ Microsoft Purview AI Shield & Deep Governance⏱ 20 min⭐ 200 BASE XP⌨ HANDS-ON LAB

Data Loss Prevention (DLP) for AI Prompts and Responses

Purview Agentic DLP inspects data in motion across two critical boundaries:

  • Inbound Prompts: Scans user input for social engineering, malicious jailbreaks, and indirect prompt injections before the prompt reaches model weights.
  • Outbound Completions: Scans LLM output for Sensitive Information Types (SITs) such as credit card numbers, national identification numbers, passwords, and private health data. If detected, the response is redacted or blocked before reaching the user interface.
⌨ HANDS-ON LABAudit Purview AI DLP Incident Logs
⭐ +200 XP

Inspect Microsoft Purview audit logs to verify that an attempted indirect prompt injection or credit card exfiltration was blocked.

1Query Purview audit log queries endpoint for recent DLP violations in Copilot sessions.
lab-sandbox — simulated environment
INFINITY LAB SANDBOX v2.6 — simulated shell
Type the command for the current objective. Helpers: "hint", "solution", "clear".
$
OBJECTIVE 1 / 1 — type "hint" if stuck
SYNAPSE VERIFICATION
QUERY 1 // 1
How does Purview Agentic DLP protect against accidental sensitive data disclosure in Copilot responses?
By blocking the internet connection
By scanning real-time generated tokens for Sensitive Information Types (SITs) and redacting or blocking matching patterns
By turning off the computer
By emailing the user's manager every time they type a prompt