How do you prevent PII and secrets from leaking into LLM prompts and logs?

Controls

• Redaction: Scan outbound prompts for emails, PAN/Aadhaar patterns, API keys — block or mask before the model call.