Safeguard your agentic AI applications with the Amazon Bedrock Guardrails InvokeGuardrailChecks API
TL;DR
AWS announced InvokeGuardrailChecks for Amazon Bedrock Guardrails on June 16, 2026. Developers can call individual safety checks inside agent loops without creating guardrail resources, IDs, or versions. The API is detect-only. It does not block, mask, or rewrite content; it returns discrete 0-to-1 scores for content filters, prompt attacks, and sensitive information.
Nauti's Take
This is a useful building block, not a safety product you can just switch on. AWS gives scores and examples; teams still own thresholds, audit trails, escalation paths, and the messy decision of what to do with ambiguous findings.
For agent systems, middle-of-loop control is more valuable than checking only the final answer. The PR spin starts when safe multi-turn agents sounds like an API call instead of an engineering discipline.
Briefingshow
Agent safety often breaks between steps: a tool returns personal data, a user attempts prompt injection, or a model continues from risky intermediate content. The API makes those points checkable without managing a separate guardrail resource for each loop stage. The tradeoff is that teams still have to design thresholds, audit logs, and response logic themselves.
Sources
