Moderations API — POST /v1/moderations with BR guardrail layer

2026-03-22

apisecurity

What We Built

OpenAI-compatible moderation endpoint at POST /v1/moderations with BR-specific guardrail categories layered on top. Returns standard OpenAI moderation format with additional br_jailbreak and br_pii_detected categories.

BR categories use pattern-based detection from existing guardrail checks (jailbreak patterns, PII regex) — no external API call needed, sub-millisecond latency.

Why It Matters

Unique in the AI gateway space — no competitor offers moderation with built-in jailbreak and PII detection as additional signal. Customers get both provider moderation AND BR's own security analysis in one API call.

Lockstep Checklist

[x] API Routes: src/api/routes/moderations.ts + mount in server.ts
[x] TS SDK: moderations.create()
[x] Python SDK: moderations.create() sync + async
[x] MCP Schemas: N/A — not agent-facing
[x] Master Record: N/A — auto-updated