Amazon Bedrock Guardrails

service

proprietary managed AWS service(no self-host)

features

denied topics, content filters, PII redaction, contextual grounding

source

closed

license

proprietary

Closed managed cloud service; no weights or source. The ApplyGuardrail API can be pointed at self-hosted models, but that is the managed guardrail reaching outward, not a self-hostable guardrail, so source stays closed.

• https://aws.amazon.com/bedrock/guardrails/ recorded 2026-08-13

  An AWS-run service with no source repository and no self-hostable implementation; use is governed by the AWS service terms rather than a published license. 'The ApplyGuardrail API allows you to use the configurable safeguards offered by Bedrock Guardrails with any foundation model whether hosted on Amazon Bedrock or self-hosted models' - the guardrail itself still runs at AWS.

AWS publishes no usage numbers for this service. The page names customers - 'Companies like Chime Financial, KONE, Panorama, Strava, Remitly, and PwC trust Bedrock Guardrails for their AI applications' - but gives no count of anything, so no numeric reach is claimed: a 100K-1M label would assert a measurement nobody made. Named customers of that kind are what level 3 rests on here, and the cited sources support it.

• https://aws.amazon.com/bedrock/guardrails/ recorded 2026-08-13

  Names six customers by logo and quote. No customer count, request volume or developer count appears on the page.

Broad configurable managed guardrail set; strong coverage, closed. Automated Reasoning checks sit as a sixth policy beside the five the value names.

• https://aws.amazon.com/bedrock/guardrails/ recorded 2026-08-13

  Page lists content moderation (content and word filters), prompt attack detection, topic classification (denied topics), PII redaction (sensitive information filters), and hallucination detection (contextual grounding and Automated Reasoning checks), applied 'consistently across any foundation model'.