AWS Weekly Roundup: Claude Mythos Preview in Amazon Bedrock, AWS Agent Registry, and more (April 13, 2026) | Amazon Web Services

Amazon Web Services (AWS) has significantly advanced its artificial intelligence (AI) capabilities and cost management tools this week, introducing crucial features for enterprise adoption of generative AI. The most impactful development is the new support for cost allocation by IAM user and role within Amazon Bedrock, a move directly addressing a critical pain point for organizations scaling AI initiatives. Complementing this is the preview release of Claude Mythos, a sophisticated AI model specifically engineered for cybersecurity tasks, and the AWS Agent Registry, designed to streamline the discovery and governance of AI agents within an organization.

Enhanced Cost Visibility for AI Workloads

The integration of IAM user and role-based cost allocation into Amazon Bedrock represents a significant step forward in providing granular financial oversight for AI deployments. As businesses increasingly move AI projects from experimentation to production, understanding resource consumption and associated costs becomes paramount for finance departments and leadership. This new feature allows customers to tag IAM principals, such as individual users or service roles, with attributes like team, project, or cost center. These tags are then activated within the AWS Billing and Cost Management console, enabling the detailed tracking of expenses.

The resultant cost data is seamlessly integrated into AWS Cost Explorer and the detailed Cost and Usage Report (CUR). This provides organizations with a clear and actionable line of sight into model inference spending. For companies deploying AI agents across multiple teams, monitoring foundation model usage by department, or utilizing specialized tools like Claude Code on Amazon Bedrock, this enhancement is poised to revolutionize how AI investments are tracked and managed. The ability to attribute specific costs to the entities responsible for their generation is vital for budgeting, resource optimization, and demonstrating the ROI of AI initiatives.

This development arrives at a time when organizations are grappling with the escalating costs associated with large-scale AI model inference. A recent industry survey indicated that over 60% of companies are increasing their AI spending in the coming year, with a significant portion citing concerns about cost management as a key challenge. The AWS solution directly addresses this by enabling proactive cost control and accountability.

Claude Mythos Preview: A New Frontier in AI-Powered Cybersecurity

Further expanding the generative AI landscape on Amazon Bedrock, AWS has launched a gated research preview of Claude Mythos, Anthropic’s most advanced AI model to date. Available through Project Glasswing, Claude Mythos is positioned as a new model class specifically designed to tackle complex cybersecurity challenges. Its capabilities include identifying sophisticated security vulnerabilities in software, analyzing extensive codebases, and delivering state-of-the-art performance in cybersecurity, coding, and intricate reasoning tasks.

This model offers a proactive approach to security by enabling security teams to discover and address potential vulnerabilities in critical software before they can be exploited by malicious actors. The preview is currently restricted to allowlisted organizations, with Anthropic and AWS prioritizing companies crucial to internet infrastructure and maintainers of open-source projects. This targeted release aims to gather critical feedback and refine the model’s capabilities in real-world, high-stakes environments.

The introduction of Claude Mythos signals a growing trend of specialized AI models tailored for specific industry verticals. As cyber threats become more sophisticated, the demand for AI solutions that can augment human expertise in threat detection, analysis, and remediation is rapidly increasing. Claude Mythos represents a significant advancement in this domain, potentially transforming how organizations approach software security.

AWS Agent Registry: Centralized Discovery and Governance for AI Agents

In parallel, AWS has introduced the AWS Agent Registry as a preview within Amazon Bedrock AgentCore. This new service provides organizations with a private catalog for the discovery, management, and governance of AI agents, tools, skills, MCP servers, and other custom resources. The primary objective of the Agent Registry is to prevent redundant development efforts by enabling teams to easily locate and leverage existing capabilities within their organization.

The registry offers robust search functionalities, including semantic and keyword-based queries, along with approval workflows to ensure proper oversight. Furthermore, it integrates with AWS CloudTrail, providing audit trails for enhanced governance and compliance. Access to the Agent Registry is available through the AgentCore Console, the AWS Command Line Interface (CLI), SDKs, and can be queried from Integrated Development Environments (IDEs) via an MCP server.

The proliferation of AI agents within enterprise environments presents both opportunities and challenges. The Agent Registry addresses the challenge of fragmentation and discoverability, ensuring that valuable AI assets are not lost or duplicated. This fosters collaboration, promotes best practices, and accelerates the deployment of AI-powered solutions by making reusable components readily accessible.

Broader Implications and Future Outlook

The confluence of these AWS announcements underscores a strategic commitment to facilitating the enterprise-wide adoption of generative AI. By addressing critical concerns around cost management, specialized model development, and organizational governance, AWS is building a comprehensive ecosystem for AI innovation.

The IAM cost allocation feature is particularly significant. As AI budgets grow, the ability to precisely attribute costs is no longer a luxury but a necessity for financial sustainability. This feature empowers finance teams to engage proactively with their engineering counterparts, fostering a shared understanding of AI’s financial footprint and enabling data-driven decisions on resource allocation and optimization.

The introduction of Claude Mythos, albeit in a preview phase, highlights AWS’s dedication to providing cutting-edge AI models that cater to specific industry needs. The cybersecurity domain is a prime candidate for AI augmentation, and a model of Mythos’s caliber could dramatically enhance an organization’s ability to defend against evolving threats.

The AWS Agent Registry, meanwhile, tackles the operational complexities of managing a growing fleet of AI agents. By creating a centralized, searchable repository, AWS is promoting efficiency and reducing the potential for duplicated effort, which is crucial for maintaining agility and controlling development costs in the fast-paced world of AI.

These developments collectively position AWS as a key enabler for organizations seeking to harness the power of AI responsibly and effectively. As the AI landscape continues to evolve at an unprecedented pace, features that enhance control, transparency, and discoverability will be instrumental in driving sustained innovation and business value. The coming months will likely see further refinements and broader availability of these features, solidifying AWS’s role in shaping the future of enterprise AI.

For detailed implementation guidance, customers can refer to the official AWS documentation. The IAM principal cost allocation documentation can be found at https://docs.aws.amazon.com/awsaccountbilling/latest/aboutv2/iam-principal-cost-allocation.html. Updates on new features and announcements are regularly published on the "What’s New with AWS" page, https://aws.amazon.com/new/.