AWS AI Updates: April 4, 2026

1. Amazon Bedrock Guardrails Adds Cross-Account Safeguards for Organization-Wide Enforcement

Amazon Bedrock. Bedrock Guardrails now supports cross-account safeguards, letting a central security team enforce content filtering and hallucination controls across every AWS account in an organization from a single management account. The new Amazon Bedrock policy type automatically applies configured guardrail IDs to organizational units and member accounts, eliminating per-account manual setup. AWS claims the filters block up to 88% of harmful multimodal content in both prompts and responses. Available in all commercial and GovCloud regions where Bedrock Guardrails is supported. Source

2. SageMaker Data Agent Adds Interactive Charting, Snowflake SQL, and Materialized View Management

Amazon SageMaker. The SageMaker Data Agent in Unified Studio notebooks now supports interactive charting via natural language prompts, cross-source SQL analytics against Snowflake tables joined with AWS Glue Data Catalog data, and automated materialized view recommendations with refresh scheduling. Users can ask the agent to analyze notebook queries and suggest optimization candidates. Available in all regions supporting SageMaker Unified Studio. Source

3. Amazon EMR Launches Spark Troubleshooting and Upgrade Agents as Kiro Powers

Amazon EMR. Two new Kiro-powered agents for Apache Spark are now available: a troubleshooting agent that analyzes logs, metrics, and configurations across EMR on EC2 and EMR Serverless to pinpoint root causes of failed jobs, and an upgrade agent that automates Spark version migrations including code transformation, dependency resolution, and data quality validation. Both integrate through MCP Proxy for AWS with IAM authentication and full CloudTrail audit logging. Available in all commercial regions. Source