AWS Weekly Update: Claude Mythos Preview, Agent Registry, Cost Allocation, and S3 Files

Introduction

In this week's AWS update, several significant announcements are reshaping how organizations manage AI workloads, security, and data storage. From enhanced cost visibility for AI services to cutting-edge AI models and improved resource discovery, AWS continues to refine its ecosystem for enterprise-scale operations. Here's a detailed look at the latest features now available or in preview.

Cost Allocation for AI-Driven Workloads

As teams accelerate their adoption of AI, finance and leadership departments often struggle to track resource usage and associated costs. Recognizing this challenge, AWS has introduced cost allocation support for IAM users and roles within Amazon Bedrock. This feature allows administrators to tag IAM principals with attributes such as team names or cost centers, then activate those tags in the Billing and Cost Management console. The resulting data flows into AWS Cost Explorer and the detailed Cost and Usage Report, providing granular visibility into model inference spending.

This capability is particularly valuable for organizations scaling agents across multiple teams, tracking foundation model usage by department, or running tools like Claude Code on Bedrock. It transforms cost tracking from a guessing game into a precise audit trail, empowering teams to optimize their AI investments. For setup instructions, refer to the IAM principal cost allocation documentation (internal anchor).

Claude Mythos Preview on Amazon Bedrock

Next-Generation AI for Cybersecurity

Anthopic's latest and most advanced AI model, Claude Mythos, is now available on Amazon Bedrock as a gated research preview under Project Glasswing. This model introduces a new class focused on cybersecurity, capable of identifying sophisticated vulnerabilities in software, analyzing large codebases, and delivering state-of-the-art performance across security, coding, and complex reasoning tasks.

Security teams can leverage Claude Mythos to discover and address critical vulnerabilities before threats emerge. Access is currently limited to allowlisted organizations, with Anthropic and AWS prioritizing internet-critical companies and open-source maintainers. This preview represents a significant leap in applying AI to proactive defense and code analysis.

AWS Agent Registry for Centralized Discovery and Governance

Streamlining AI Agent Management

AWS has launched the Agent Registry through Amazon Bedrock AgentCore, now in preview. This service provides organizations with a private catalog for discovering and managing AI agents, tools, skills, MCP servers, and custom resources. It helps teams locate existing capabilities rather than duplicating them, reducing redundancy and improving efficiency.

Key features include:

• Semantic and keyword search for finding relevant agents
• Approval workflows to govern access and usage
• CloudTrail audit trails for compliance and monitoring
• Accessible via AgentCore Console, AWS CLI, SDK, and as an MCP server queryable from IDEs

This registry is a foundation for enterprise AI governance, ensuring that teams can reuse and scale their AI investments responsibly. For more details, see the Agent Registry documentation (internal anchor).

Amazon S3 Files: S3 Buckets as Shared File Systems

Bridging Storage and Compute

AWS announced Amazon S3 Files, a new capability that transforms S3 buckets into shared file systems. Built on Amazon EFS technology, it connects any AWS compute resource directly with S3 data, delivering full file system semantics with low-latency performance. Caching frequently used data, it can provide aggregate throughput of multiple terabytes per second.

This is ideal for applications that require file-based access patterns but need the scalability and durability of S3. Use cases include data lakes, machine learning preprocessing, and high-performance computing. S3 Files eliminates the need to move data between storage and compute, simplifying workflows and reducing costs. For best practices, refer to the S3 Files documentation (internal anchor).

Conclusion

This week's updates underscore AWS's commitment to making AI more manageable, secure, and cost-effective. From granular cost allocation to the powerful Claude Mythos model, the new Agent Registry, and innovative S3 Files, these features provide practical tools for scaling AI adoption. For developers and IT leaders, now is the time to explore these previews and prepare for production use. Stay tuned for more advancements as AWS continues to evolve its platform.