PII Redaction + Guardrails
The legal/security shape — Presidio-based PII detection at the gateway, prompt-injection detection inbound, content filtering outbound, and per-key guardrail policies.
💡
Tip: Guardrails belong at the gateway, not in every application. One enforcement point, one audit trail, one place to update policy.
What this use case covers
- PII detection and redaction with Microsoft Presidio at the gateway boundary — emails, phone numbers, SSNs, credit cards.
- Prompt-injection detection on inbound requests, before they reach the model.
- Content filtering on outbound responses, with a configurable policy.
- Per-key guardrail policies — some teams need stricter rules (regulated workloads) than others (internal tools).
- A brief compliance-posture note covering SOC2, HIPAA, and GDPR concerns and where the gateway helps vs where it does not.