AWS News - 2026-11-07

AWS Data Exports for Cost and Usage Report (CUR) 2.0 and AWS Cost Explorer now amortize the upfront fee for Amazon EC2 Capacity Blocks for ML across the duration of each reservation. You can see the amortized cost of each instance launched into a Capacity Block and the amortized cost of unused capacity, so you can attribute your upfront fee to the workloads that used it and identify capacity that went unused.

Previously, a Capacity Block's upfront fee appeared as a single charge on the purchase date, making it difficult to see how that cost was spread across usage over the reservation. Now the amortized cost appears alongside your instance usage, so you can allocate GPU costs to the teams, projects, and applications that ran on the reserved capacity. You can also see what share of your reservation went unused and plan future Capacity Block purchases. In CUR 2.0, the amortized costs appear in the reservation_amortized_upfront_fee_for_billing_period, reservation_effective_cost columns, and reservation_net_effective_cost shows the amortized cost after discounts. In Cost Explorer, select the amortized cost view to see Capacity Block costs allocated for the reservation.

 Amortization for Capacity Blocks is available at no additional charge in all AWS Regions except the AWS GovCloud (US) Regions and the China Regions. To learn more, see the CUR 2.0 data dictionary and exploring your data using Cost Explorer.

You can now use Amazon MSK Replicator in the AWS GovCloud (US-East, US-West) Regions to replicate streaming data across Amazon MSK clusters and from external Apache Kafka clusters including on-premises, self-managed on AWS, or other cloud providers to Amazon MSK Provisioned clusters.

MSK Replicator is a feature of Amazon MSK that automates data replication between Kafka clusters, eliminating the need to manage custom replication infrastructure or configure open-source tools. With MSK Replicator, you can easily build regionally resilient streaming applications for increased availability and business continuity. MSK Replicator provides automatic asynchronous replication across Kafka clusters, eliminating the need to write custom code, manage infrastructure, or setup cross-region networking. MSK Replicator automatically scales the underlying resources so that you can replicate data on-demand without having to monitor or scale capacity. MSK Replicator also replicates the necessary Kafka metadata including topic configurations, Access Control Lists (ACLs), and consumer group offsets. If an unexpected event occurs in a region, you can failover to the other AWS Region and seamlessly resume processing.

You can get started with MSK Replicator from the Amazon MSK console or the Amazon CLI. To learn more, visit the MSK Replicator documentation, product page, and pricing page.

Amazon Managed Workflows for Apache Airflow (Amazon MWAA) Serverless now supports Apache Airflow connections and variables, so your workflows can reach databases, APIs, and other external systems. You can manage credentials and configuration outside your workflow code.

You store sensitive connections and variables in the secrets backend of your choice, such as AWS Secrets Manager, AWS Systems Manager Parameter Store, or your own custom backend, configure it on your workflow, and grant your workflow execution role access. Tasks retrieve them at run time through standard Apache Airflow methods. You can define non-secret values as environment variables on the workflow.

This feature is available in all AWS Regions where Amazon MWAA Serverless is available. For more information, see Amazon MWAA Serverless Regions.

To learn more, see Managing connections and variables and Using environment variables in Amazon MWAA Serverless workflows in the Amazon MWAA Serverless User Guide.

Amazon MWAA Serverless now supports operators from Apache Airflow provider packages, such as HTTP, Snowflake, or dbt, and custom operators that you write. This gives data engineers the flexibility to integrate with more systems and build specialized workflow tasks without provisioning infrastructure.

To get started, add the provider package or your own operator to a code package, upload it to Amazon S3, and reference the operator in your workflow definition when you create or update a workflow.

This feature is available in all AWS Regions where Amazon MWAA Serverless is available. For a list of supported Regions, see Amazon MWAA Serverless Regions. To learn more, see Running custom code with Python, Bash, and custom operators in the Amazon MWAA Serverless User Guide.

AWS Agent Registry, available through Amazon Bedrock AgentCore, is now available in the AWS GovCloud (US-West) Region.

AWS Agent Registry is a fully managed discovery service that gives your organization a central catalog for organizing, curating, and discovering resources. You can publish MCP servers, tools, agents, agent skills, and custom resources to a searchable registry, and control what becomes discoverable through an approval workflow. People and AI agents can then find the right tools and agents through hybrid search, catalog browsing, and a native MCP endpoint.

With Agent Registry in AWS GovCloud (US-West), public sector organizations and their partners get a single governed inventory of the agents and tools in their environment. Approval workflows and AWS CloudTrail audit logs support their compliance and oversight requirements.

To get started, see the AWS Agent Registry documentation.

Today, Kimi K3.1 Pro from Moonshot AI is available on Amazon Bedrock, expanding our open weight model portfolio with the security and governance that customers rely on. Customers using Amazon Bedrock can now run agentic use cases such as coding and document processing on Kimi K3.1 Pro (Highspeed), which delivers high output tokens per second (OTPS) for faster, more responsive results.

According to Moonshot AI, Kimi K3.1 Pro is an open-weight, natively multimodal agentic model with a 1-million-token context window, and its most capable 3-trillion-parameter-class model to date. Moonshot AI reports improvements in coding, 3D creation, financial research, and other agentic work. On Amazon Bedrock, it runs within the same security boundary as proprietary models and supports explicit prompt caching to reduce latency and input costs.

Kimi K3.1 Pro is available in all AWS Regions where Amazon Bedrock is available through cross-Region inference. To get started, visit the Amazon Bedrock console. To learn more, see the Amazon Bedrock documentation.

Auto は裏側でモデルカスケードを使っています。どのセッションも効率のよい最初のモデルで始まり、独立したチェックがバックグラウンドで作業を確かめ、必要になったときに Auto はセッションをより高性能なモデルに引き継ぎます。

Protocols like Android Debug Bridge (ADB) and Web Real-Time Communication (WebRTC), by default, use unencrypted channels for network communications and don’t support enterprise authentication mechanisms like JSON Web Tokens (JWTs). In a multi-tenant Kubernetes cluster—including Amazon Elastic Kubernetes Service (Amazon EKS)—pods can reach other tenants’ open ports by default unless you configure network policies to […]

Bulletin ID: 2026-173-AWS
Publication Date: 11/06/2026 09:00 AM PST

Description:

We identified CVE-2026-122193, an OS command injection issue in an example provided in the set-up documentation. The example included a script that creates AWS Systems Manager (SSM) documents (Rex-RO and Rex-RW) for running scripts on managed instances. Due to how the example script handled input parameters, a user with permissions to invoke the SSM document could supply crafted input that would run unintended operating system commands on the managed instance, outside of the REX sandbox. The scriptArgs parameter is affected by the same class of issue.

Impacted versions:

This is not an issue in a shipped REX artifact version.

Refer to the article below for the most up-to-date and complete information.

Read this article

Bulletin ID: 2026-166-AWS
Publication Date: 11/06/2026 10:00 AM PST

Description:

We identified two issues in the Babelfish TDS protocol layer:
- CVE-2026-119923: A memory-safety issue in the TDS connection-handshake handling could allow a remote, unauthenticated actor to affect the database process.
- CVE-2026-119924: A memory-safety issue in the TDS remote-procedure-call request handling could allow a remote, authenticated actor to read adjacent process memory or cause a denial of service.

Refer to the article below for the most up-to-date and complete information.

Read this article

Run Claude Code as a remote agent on Amazon Bedrock AgentCore. Each coding task runs in its own isolated session in your AWS account, every model call goes through Amazon Bedrock, and the result comes back as a GitHub pull request you review and merge.

Taxbit built Taxbit Assist, a cited, permission-aware AI assistant embedded in its enterprise dashboard, on Amazon Bedrock AgentCore. This post shows how they grounded answers in a curated knowledge base, enforced role-aware and region-compliant safety, and continuously measured answer quality on live production traffic.

Greenlight Financial resolves more than 90% of up to 4,000 monthly security and risk cases without human intervention using Luna, a tiered multi-agent system on Amazon Bedrock AgentCore. This post covers the architecture, episodic memory for continuous learning, and the expansion from SOC operations into financial risk triage.

An AI agent that acts for a user raises an identity question that per-app OAuth and service accounts leave open: which agent may act for which user, and how you stop it. In this first post of two, you build a task agent on Amazon Bedrock AgentCore that governs access through Okta Cross App Access and AgentCore Identity, and enforces each user's authority.

In Part 1 you built one governed agent. Part 2 scales the pattern to an enterprise: Amazon Bedrock AgentCore Gateway becomes the single front door and redeems Okta's grant, Cedar policies bind each tool to the agent, a second agent runs on Amazon EKS, and AgentCore Observability traces every allow and deny beside Okta's log in Amazon CloudWatch.

Learn how to build a serverless Model Context Protocol (MCP) server on Amazon ECS with AWS Fargate that gives AI agents secure, metadata-driven access to an enterprise data lake. Exposing a new dataset requires only a JSON file, with no per-dataset integration code.

As we built Amazon Quick, its agents grew from answering questions to taking actions, and the same primitives kept getting rebuilt under different names. This post shares the taxonomy we converged on — triggers, agents, capabilities, and learned context — found by decomposing agentic work into four plain questions: When, What, Who, and How.

As we built Amazon Quick, its agents grew from answering questions to taking actions, and the same primitives kept getting rebuilt under different names. This post shares the taxonomy we converged on — triggers, agents, capabilities, and learned context — found by decomposing agentic work into four plain questions: When, What, Who, and How.

Amazon SageMaker Inference now supports GPU partitioning with NVIDIA Multi-Instance GPU (MIG), which divides one GPU into hardware-isolated partitions so you can run a different model in each or take a right-sized fraction of a newer accelerator. This post shows how two optional API fields place one model per partition, and what we measured against a whole older GPU.