AWS and Anthropic Launch Claude Mythos Cybersecurity Model, New AI Cost Governance Tools
AWS and Anthropic Launch Claude Mythos Cybersecurity Model, New AI Cost Governance Tools
AWS and Anthropic today unveiled Claude Mythos, a new AI model class specifically designed for cybersecurity, now available as a gated research preview on Amazon Bedrock through Project Glasswing. The model can identify sophisticated vulnerabilities in software, analyze large codebases, and perform complex reasoning tasks, aiming to help security teams find threats before they emerge.
Alongside the model launch, AWS introduced IAM principal cost allocation for Amazon Bedrock, enabling teams to tag AI resource usage by user, team, or cost center. The data flows into AWS Cost Explorer and detailed Cost and Usage Reports, providing granular visibility into model inference spending.
“Claude Mythos represents a leap forward in AI-driven security,” said an Anthropic spokesperson. “By combining advanced reasoning with deep code analysis, it can uncover vulnerabilities that often escape traditional tools.”
The cost allocation feature addresses a growing pain point for enterprises scaling AI. “Our customers have told us they need to track AI spend by department and project,” said an AWS executive. “This new capability gives finance and leadership the clarity they need to govern AI investments effectively.”
AWS Agent Registry Now in Preview
AWS also launched Agent Registry through Amazon Bedrock AgentCore, providing a private catalog for discovering and managing AI agents, tools, skills, MCP servers, and custom resources. The registry includes semantic and keyword search, approval workflows, and CloudTrail audit trails, helping teams avoid duplicating existing capabilities. It is accessible via the AgentCore Console, AWS CLI, SDK, and as an MCP server queryable from IDEs.
Amazon S3 Files Makes Storage Accessible as File Systems
Another notable launch is Amazon S3 Files, which converts S3 buckets into shared file systems with full file system semantics and low latency performance. Built on Amazon EFS technology, it supports multiple terabytes per second of aggregate throughput, enabling direct connection between compute resources and S3 data.
Background
The announcements come amid rapid AI adoption where organizations are moving from experimentation to production. A common theme in recent customer workshops has been the need for better cost visibility and governance across AI tools. The new cost allocation feature directly responds to these demands by allowing IAM principals to be tagged with attributes like team or cost center, with tags activated in Billing and Cost Management.
Claude Mythos addresses a parallel need for specialized AI in cybersecurity. As software supply chains grow more complex, organizations require advanced tools to proactively identify vulnerabilities. The model is currently limited to allowlisted organizations, with Anthropic and AWS prioritizing internet-critical companies and open source maintainers.
What This Means
For enterprises, the combination of cost allocation and agent governance provides a framework for scaling AI responsibly. Finance teams can now attribute model inference costs to specific business units, while engineering teams can discover and reuse existing AI agents rather than rebuilding them. This reduces redundancy and operational overhead.
The Claude Mythos preview signals a new frontier for AI in cybersecurity. If successful, it could shift how organizations approach vulnerability discovery—from periodic scans to continuous, AI-driven analysis. The partnership between AWS and Anthropic also underscores the trend of integrating advanced AI models directly into cloud platforms, making them accessible to a wide range of customers.
These updates are available now through AWS’s regular channels. For more details on cost allocation setup, refer to the Background section above and the official AWS documentation.