Amazon Bedrock AgentCore now supports Guardrails in policy, evaluating outputs of authorized agent actions and tool calls in real-time to detect and block prompt injection attacks, harmful content, and sensitive information exposure. Policy evaluations occur at the gateway perimeter for consistent enforcement regardless of agent autonomy.
Amazon Bedrock AgentCore - Guardrails in Policy
calendar_today
June 17, 2026
domain
amazon-elasticache