Amazon Bedrock Guardrails now supports cross-account safeguards in general availability, allowing organizations to centrally enforce AI safety controls across all AWS accounts within an AWS Organization. A single guardrail can be defined in the management account and automatically applied to every model invocation across all member accounts and organizational units. The feature supports both organization-level and account-level enforcement, with options to include or exclude specific models and apply comprehensive or selective content guarding for system and user prompts. Key limitations include no support for Automated Reasoning checks, and incorrect guardrail ARNs can block model access entirely.
Sort: