AWS has launched a gated preview of Anthropic's most advanced AI model, Claude Mythos, on Amazon Bedrock, alongside a new Agent Registry for centralized AI agent governance. The announcements, made this week, signal a major push to address enterprise security and operational efficiency.
Claude Mythos introduces a new model class dedicated to cybersecurity. It can identify sophisticated vulnerabilities in software, analyze large codebases, and deliver state-of-the-art performance across security, coding, and complex reasoning tasks. Access is currently limited to allowlisted organizations, prioritizing internet-critical companies and open source maintainers.
"Claude Mythos represents a leap forward in using AI for proactive defense," said an Anthropic spokesperson. "It helps security teams discover and fix vulnerabilities before they become threats."
Meanwhile, AWS Agent Registry is now in preview via Amazon Bedrock AgentCore. It provides a private catalog for discovering and managing AI agents, tools, skills, MCP servers, and custom resources. The registry helps teams avoid duplicating existing capabilities through semantic and keyword search, approval workflows, and CloudTrail audit trails.
"The Agent Registry prevents reinventing the wheel," stated an AWS executive. "It is accessible via the AgentCore Console, AWS CLI, SDK, and as an MCP server queryable from IDEs."
AWS also introduced cost allocation support for Amazon Bedrock by IAM user and role. This allows organizations to tag IAM principals with attributes like team or cost center, activate them in Billing and Cost Management, and track model inference spending via Cost Explorer and the Cost and Usage Report. The feature addresses a common pain point for teams scaling AI from experimentation to production.
"Better cost visibility is critical as AI adoption accelerates," noted an AWS product manager. "Finance and leadership need to know who is using which resources and at what cost."
Background
In recent years, enterprises have rapidly adopted AI-driven workflows, but many struggle with managing costs and security risks. AWS's latest updates directly tackle these challenges. Claude Mythos offers a specialized AI for cybersecurity, anticipating a rise in sophisticated attacks. The Agent Registry simplifies governance across growing agent ecosystems.
The cost allocation feature emerges from feedback during AI-Driven Development Lifecycle (AI-DLC) workshops, where teams emphasized the need for granular spend tracking. By integrating with existing AWS billing tools, Bedrock now provides clear cost attribution for model inference.
What This Means
Claude Mythos enables organizations to augment security teams with AI that can scan entire codebases for vulnerabilities. This could shift security testing left in the development lifecycle, reducing the window for exploits. The gated preview ensures responsible deployment, with Anthropic focusing on high-impact users first.
The Agent Registry reduces redundancy and improves operational efficiency. Teams can quickly find and reuse existing AI agents, speeding up development and reducing costs. Combined with audit trails, it strengthens compliance and oversight.
Cost allocation empowers finance departments to accurately bill internal teams for AI usage, encouraging responsible consumption. This aligns with AWS's broader push to make generative AI more enterprise-ready.
Additionally, Amazon S3 Files was announced, making S3 buckets accessible as shared file systems with full semantics and low-latency performance. While separate, it complements the AI infrastructure by simplifying data access for compute resources.
For more details, see the official IAM principal cost allocation documentation and the Agent Registry setup guide.