Amazon Bedrock Guardrails adds API for granular agentic AI workflow control
Amazon Bedrock Guardrails now offers the InvokeGuardrailChecks API, enabling granular, per-request control over safeguards in agentic AI applications. This new resourceless API allows developers to apply individual content, prompt attack, and sensitive information filters at any point in an agent's loop, returning severity and confidence scores for custom actions. It addresses the challenge of scaling guardrails in iterative AI workflows by operating in a detect-only mode, offering flexibility for evolving requirements. The API is available today in select AWS Regions.
- →New API for granular safeguard application in agentic AI workflows
- →Supports multiple safeguard types including prompt attack detection
- →Flexible scoring and custom action enablement
- →Availability in select AWS Regions
Features (2) ›
- New API for granular safeguard application in agentic AI workflows
Amazon Bedrock Guardrails introduces the InvokeGuardrailChecks API, a resourceless API that allows applying individual safeguards at any point in agentic AI applications. This provides granular, per-request control over which safeguards to run at each step of an agent's loop.
- Supports multiple safeguard types including prompt attack detection
The API supports content filters, prompt attack detection (jailbreak, prompt injection, prompt leakage as independent checks), and sensitive information filters for PII. Prompt attack detection is exposed as a separate safeguard for granular invocation.
Enhancements (1) ›
- Flexible scoring and custom action enablement
The API returns numeric severity and confidence scores, enabling developers to implement custom thresholds and actions such as blocking, passing, retrying, or logging based on specific requirements. This supports the iterative nature of agentic AI applications where each step has a different risk profile.
Notes (1) ›
- Availability in select AWS Regions
The InvokeGuardrailChecks API is available in US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (London), Europe (Stockholm), Asia Pacific (Tokyo), and Asia Pacific (Sydney).
https://aws.amazon.com/about-aws/whats-new/2026/06/amazon-bedrock-guardrails-api-ai/