Amazon Web Services (AWS) has unveiled a series of strategic updates designed to tighten the intersection of generative AI, cybersecurity, and cloud infrastructure management. Chief among these is the introduction of the Claude Mythos Preview in Amazon Bedrock, a specialized model class from Anthropic tailored specifically for the high-stakes demands of cybersecurity professionals.
This rollout comes at a critical time as enterprises transition from AI experimentation to full-scale production. To address the financial complexities of this shift, AWS has also launched new support for cost allocation by IAM user and role within Amazon Bedrock. This allows organizations to tag IAM principals with specific attributes, such as cost centers or team names, providing a granular view of model inference spending through the AWS Cost Explorer and detailed Cost and Usage Reports.
Beyond AI, the cloud provider is expanding its hardware and data access capabilities, notably with the addition of Rigetti’s 108-qubit Cepheus QPU to Amazon Braket and the launch of Amazon S3 Files. Together, these updates signal a push toward more specialized AI applications and a more flexible approach to how data is accessed across compute resources.
Claude Mythos and the New Frontier of AI Cybersecurity
The arrival of Claude Mythos on Amazon Bedrock marks a significant pivot toward domain-specific AI. Available as a gated research preview through Project Glasswing, Claude Mythos is described as Anthropic’s most sophisticated model to date, specifically engineered to identify complex security vulnerabilities within software and analyze massive codebases. AWS Weekly Roundup (April 13, 2026).
For security teams, the utility of this model lies in its ability to perform complex reasoning tasks and coding analysis to discover vulnerabilities before they can be exploited by external threats. Due to the fact that of the sensitivity of these capabilities, access is currently restricted to allowlisted organizations. AWS and Anthropic are reportedly prioritizing open-source maintainers and companies deemed critical to internet infrastructure.
Centralizing AI Governance with AWS Agent Registry
As companies deploy an increasing number of AI agents, the risk of redundant development and fragmented governance grows. To combat this, AWS has introduced the Agent Registry through Amazon Bedrock AgentCore. This service functions as a private catalog where organizations can discover and manage AI agents, tools, skills, and MCP (Model Context Protocol) servers.
The registry is designed to prevent the duplication of capabilities by allowing teams to locate existing tools via semantic and keyword search. To maintain enterprise-grade control, the system includes approval workflows and CloudTrail audit trails. Developers can interact with the registry through the AgentCore Console, AWS CLI, SDK, or as an MCP server queryable directly from integrated development environments (IDEs).
Infrastructure Evolution: S3 Files and Quantum Scaling
AWS is also redefining how data interacts with compute resources. The new Amazon S3 Files transforms S3 buckets into shared file systems, leveraging Amazon EFS technology to provide low-latency performance and full file system semantics. This allows applications to access S3 data through both file system and S3 APIs simultaneously, eliminating the need for extensive code modifications or data migration.
In the realm of quantum computing, Amazon Braket has expanded its hardware offerings by adding support for Rigetti’s Cepheus-1-108Q device. This represents the first superconducting quantum processor on the platform to exceed 100 qubits. The device utilizes a modular design with twelve 9-qubit chiplets and CZ gates to improve resilience against phase errors. Researchers can access the QPU via several frameworks, including the Braket SDK, Qiskit, CUDA-Q, and Pennylane.
Observability and Administrative AI Tools
To assist with the operational overhead of these complex systems, AWS has released two key AI-driven tools:
- Amazon OpenSearch Service Updates: Now supports Managed Prometheus and agent tracing, consolidating metrics, logs, and AI agent tracing into a single interface. This allows operations teams to correlate slow traces with logs and use PromQL for direct query support.
- Amazon WorkSpaces Advisor: A generative AI-powered tool designed for IT administrators to troubleshoot Amazon WorkSpaces Personal deployments. The advisor automatically detects configuration problems and provides actionable recommendations to restore service.
Summary of Recent AWS Developments
| Feature/Service | Primary Function | Availability Status |
|---|---|---|
| Claude Mythos | Cybersecurity-focused AI model | Gated Research Preview |
| Agent Registry | Centralized AI agent discovery | Preview |
| Amazon S3 Files | S3 buckets as shared file systems | Available |
| Rigetti Cepheus QPU | 108-qubit quantum processing | Available via Braket |
| Bedrock Cost Allocation | IAM-based cost tracking | Available |
Looking ahead, AWS will host a virtual event titled “What’s Next with AWS” on April 28 at 9 a.m. PT. The livestream will feature AWS CEO Matt Garman, SVP Colleen Aubrey, and leaders from OpenAI to discuss the transformation of business operations through agentic AI and emerging platform capabilities.
We invite our readers to share their thoughts on these updates in the comments below. Which of these AI-driven tools do you believe will have the most immediate impact on enterprise security?
Related reading