AWS Security Digest·Week 10 of 2026·Mar 3-9, 2026·4 items

    AgentCore Picks Up Cedar Policies

    Bedrock AgentCore Policy hits GA, mixing LLM authorship with Cedar policy-as-code, the first AWS service to do that at scale. IAM gets a simplified role-creation flow with inline panels. AWS adds DESC 2026 certification for the UAE.

    In this issue1high1medium2info

    Highlights

    4 items
    $ tail -f /var/log/aws-security.log
    high/Feature Launch/

    Amazon Bedrock AgentCore Policy GA - Cedar-Based Agent Controls

    Centralized, fine-grained controls for agent-to-tool interactions are now generally available. Policies are authored in natural language and converted to Cedar (the open-source authorization language by AWS). Operates outside agent code, enabling security teams to constrain AI agent behavior without modifying application logic. Available in 13 AWS Regions.

    BedrockCedar
    info/Service Update/

    AWS Simplifies IAM Role Creation in Service Workflows

    New inline panel for creating and customizing IAM roles directly within service workflows (EC2, Lambda, EKS, ECS, Glue, CloudFormation, etc.) without switching to the IAM console. Reduces context-switching and helps developers create roles with appropriate permissions from within their service context.

    IAM
    info/Compliance/

    AWS Completes DESC 2026 Certification Audit (UAE)

    AWS renewed Tier 1 CSP certification from Dubai Electronic Security Centre for the Middle East (UAE) Region. Valid to January 22, 2027. Now covers 108 services (10% increase from previous year). Validated by BSI.

    Compliance
    medium/Threat Intel/

    February AWS Permissions Recap: GenAI Supply Chain Risk

    Security Boulevard analysis showed February's AWS permission expansion pivoted from core infrastructure to GenAI supply chain, with new fine-tuning capabilities in the Bedrock ecosystem shifting risk from data access to model behavior influence. Worth reviewing if you manage IAM policies for ML teams.

    IAMBedrock

    Key Takeaway

    1 item
    $ cat WEEKLY_SUMMARY.md

    Bedrock AgentCore Policy is a significant security primitive for the AI agent era. Instead of trusting agents to behave, you define constraints in Cedar that operate outside the agent's code. This is defense in depth applied to AI - the agent cannot circumvent policies it does not control.

    Filed Under
    BedrockAgentCoreCedarIAMDESCComplianceGenAI

    Need Custom Security Briefings?

    These weekly digests are a starting point. Contact us for tailored threat briefings, security assessments, and architectural guidance for your AWS environment.