aws AWS Security Blog ·

Enforcing User Authorization Context in Amazon Bedrock AI Agents

blogsecurityawsarchitectaws-dynamodbaws-bedrock
announcement

AWS details a pattern for propagating user authorization context within AI agents built with Amazon Bedrock AgentCore. This approach ensures agents only access data users are authorized to see, preventing unauthorized data exposure and protecting against prompt injection. It employs Amazon Cognito, Bedrock AgentCore Identity, AWS IAM, and session tags to enforce least privilege across services like DynamoDB, S3-backed Knowledge Bases, and external SaaS platforms like Salesforce. The architecture emphasizes the agent as an orchestrator, shifting authorization enforcement to downstream services.

Notes (1)
  • Pattern for User Authorization Context in Bedrock AI Agents

    AWS introduces a best practice pattern for securely propagating user authorization context in AI agents built with Amazon Bedrock AgentCore. This architecture uses Amazon Cognito and AWS STS session tags to enforce least privilege access to various data sources like DynamoDB, S3-backed Knowledge Bases, and external services such as Salesforce, without requiring authorization logic within the agent itself.

Read the original announcement →

https://aws.amazon.com/blogs/security/propagate-user-authorization-context-in-ai-agents-with-amazon-bedrock-agentcore/

Related releases