Amazon Bedrock AgentCore adds Guardrails for AI agent safety
Amazon Bedrock AgentCore now integrates with Bedrock Guardrails, enhancing safety and security for AI agents in production by providing real-time defense against prompt injection and data exposure. This feature offers centralized enforcement outside agent code, with logging via AgentCore observability, and is available in select AWS regions. It requires no new infrastructure and uses consumption-based pricing for policy evaluations.
- →Bedrock AgentCore integrates Bedrock Guardrails for enhanced AI agent safety
- →Guardrail enforcement is external to agent code for consistent policy application
- →New feature requires no additional infrastructure and offers flexible policy authoring
- →Bedrock Guardrails in policy are available in multiple AWS regions
Features (1) ›
- Bedrock AgentCore integrates Bedrock Guardrails for enhanced AI agent safety
Amazon Bedrock AgentCore now supports Bedrock Guardrails within its policy capabilities, offering enterprises deeper safety and security controls for AI agents. Guardrails evaluate agent actions and gateway inputs in real-time to detect and block prompt injection, harmful content, and sensitive data exposure.
Enhancements (1) ›
- Guardrail enforcement is external to agent code for consistent policy application
Guardrail results are evaluated at the AgentCore gateway perimeter, outside the agent's code, ensuring consistent enforcement regardless of the agent's autonomy. All policy evaluations are logged via AgentCore observability for optimization and auditing.
Notes (2) ›
- New feature requires no additional infrastructure and offers flexible policy authoring
AgentCore policy works with existing AgentCore gateway deployments, requiring no new infrastructure. Customers can author policies using natural language or policy-as-code, with consumption-based pricing for policy evaluations.
- Bedrock Guardrails in policy are available in multiple AWS regions
This new capability is currently available in US East (N. Virginia), Europe (London), Europe (Stockholm), Asia Pacific (Sydney), and Asia Pacific (Tokyo).
https://aws.amazon.com/about-aws/whats-new/2026/06/amazon-bedrock-agentcore-policy-guardrails-generally-available