AWS News - 2026-09-11

Amazon Redshift RG instances, powered by AWS Graviton processors, are now available in the AWS Europe (Zurich) Region. RG instances deliver better performance, running data warehouse and data lake workloads up to 2.4x as fast as previous generation RA3 instances, at 30% lower price per vCPU. RG instances include Redshift's custom-built vectorized data lake query engine that processes Apache Iceberg and Parquet data on your cluster nodes, enabling you to run SQL analytics across your data warehouse and data lake using a single engine.

RG instances are available in four instance sizes, rg.large, rg.xlarge, rg.4xlarge and rg.12xlarge. Customers with existing RA3 clusters can upgrade them to RG using Snapshot & Restore, Elastic Resize, or Classic Resize. RG instances are available with flexible pricing options, including On-Demand, and 1-year and 3-year Reserved Instances with All Upfront, Partial Upfront, and No Upfront payment options. For pricing details, visit the Amazon Redshift pricing page.

To get started, refer to the following resources:

• Amazon Redshift RG Instance Documentation

• RA3 to RG Upgrade Guide

• Amazon Redshift pricing page

Amazon Elastic Container Service (Amazon ECS) now supports the IAM condition keys for CPU and memory resources on the RunTask and StartTask APIs. Administrators can use these keys to enforce consistent CPU and memory limits across all methods of launching ECS tasks. This helps organizations prevent unexpected cost overruns and keep workloads aligned with their resource policies.
Previously, the ecs:task-cpu and ecs:task-memory condition keys were available only on the RegisterTaskDefinition, CreateService, and UpdateService APIs. These condition keys are now extended on the RunTask and StartTask APIs. Now, IAM policies that reference these condition keys are evaluated when tasks are launched through RunTask and StartTask as well, giving administrators a single, unified mechanism to control resource allocation across their ECS environments.

This enhancement is available in all AWS Regions where Amazon ECS is available, at no additional cost. To learn more about using condition keys with Amazon ECS, refer to our documentation.

Today, AWS announces the general availability of second-generation single-rack AWS Outposts, a self-contained 42U rack that integrates compute, storage and networking into a single compact unit purpose-built for workloads requiring low latency, local data processing, and data residency in space and power constrained locations. A single-rack Outposts delivers up to 2,688 vCPU and 100 TB of Amazon Elastic Block Store (Amazon EBS) storage. Moreover, like multi-rack Outposts, single-rack Outposts support the latest x86-powered EC2 instances, including general purpose (M7i, M8i), compute-optimized (C7i, C8i), memory-optimized (R7i, R8i), and Outposts accelerated networking (Bmn-sf2e, Bmn-cx2, Bmn-cx3a) instances.

For organizations that operate in locations with limited rack space, such as manufacturing, gaming, and other industries, single-rack Outposts brings the latest AWS compute, storage, and networking features on-premises, and gives customers a direct path to modernize while leveraging their currently available space and power. Single-rack and multi-rack Outposts offer customers a consistent experience with the same AWS APIs, management console, automation, governance policies, and security controls across AWS Regions and on-premises locations.

For a current list of AWS Regions and countries/territories where Outposts racks are supported, check out the Outposts rack FAQs page. To get started, open the AWS Outposts console.

Amazon MQ now supports RabbitMQ version 4.3 which adds quorum queue feature enhancements such as compaction, increased priority levels, native delayed retries, and graceful consumer timeouts. RabbitMQ 4.3 also includes various bug fixes and performance improvements for memory management.

Quorum queues on RabbitMQ 4.3 performs compaction to reduce disk usage for queues and native support for 32 strict priority levels, compared to the relative 2 levels supported in previous RabbitMQ versions. Quorum queues can now automatically set failed messages aside and retry delivery after a set cooldown delay. Consumer timeouts have moved from global protocol channels to quorum queues and can be configured specific to the protocol now. Both consumer timeouts and delayed retries can be configured and managed by RabbitMQ Policies. Transient non-exclusive queues, Global QoS, and Classic queues v1 storage are no longer supported on RabbitMQ 4.3. Consumer timeouts also do not apply to classic queues.

To start using RabbitMQ 4.3 on Amazon MQ, simply select RabbitMQ 4.3 when creating a new broker using the m7g instance type through the AWS Management console, AWS CLI, or AWS SDKs. Amazon MQ automatically manages patch version upgrades for your RabbitMQ 4.3 brokers, so you need to only specify the major.minor version. To learn more about the changes in RabbitMQ 4.3, see the Amazon MQ release notes and the Amazon MQ developer guide. This version is available in all regions where Amazon MQ m7g type instances are available today. 

Today, AWS Lambda durable functions announces an integration with Pydantic AI, an open source framework for building AI agents in Python. AWS Lambda durable functions saves your Pydantic AI agent's progress as it runs, so after an interruption like a timeout, your agent resumes from the last completed step instead of starting over. Your agent gains fault tolerance without you having to write the checkpoint and retry logic yourself.

With this integration, each model and tool call your agent makes is a durable execution step, so an interrupted run does not repeat calls that already completed. This matters when the work is expensive to repeat, such as a chain of model calls that reviews a set of documents or researches a topic across many sources, where starting over means paying again for tokens to do the same work. It also helps to avoid unwanted side-effects when resuming execution, such as billing a customer twice. Because your agent runs on AWS Lambda, you manage no servers and pay only for the compute it uses.

You can use this integration in any Python AWS Lambda durable function. It is available in all AWS Regions where AWS Lambda durable functions is available. To get started, install Pydantic AI and follow its AWS Lambda durability page. You can also find the integration details in the durable execution SDK reference. For more information about AWS Lambda durable functions, see the developer guide and the AWS Lambda product page.

AWS Lambda recursive loop detection is now supported for functions running in Europe Sovereign Cloud. Recursive loop detection automatically detects and stops recursive invocations between Lambda functions and other supported services, preventing unexpected billing caused by unintended recursive loops.

Customers use event sources such as Amazon S3, Amazon SQS, and Amazon SNS to build event driven applications that trigger Lambda functions. Misconfiguration or code defect can cause events to be sent back to the same source that triggered the Lambda function, causing recursive loops and unintended usage. When such a loop is detected, recursive loop detection automatically stops processing the event and sends you an AWS Health Dashboard notification with troubleshooting steps.

Recursive loop detection is enabled by default for Lambda functions using a supported SDK version. If your function intentionally uses recursive loops, you can use the PutFunctionRecursionConfig API to turn off recursive loop detection on your Lambda function. 

To learn about recursive loop detection, visit Lambda documentation.

Amazon API Gateway now supports configurable delivery destinations and larger log events for REST API execution logs. Previously, execution logs were delivered to a single API Gateway-managed CloudWatch Logs log group with log events truncated at 1 KB, limiting visibility into request and response data.

You can now route execution logs up to 1 MB to your own Amazon CloudWatch Logs log groups, Amazon S3 buckets, or Amazon Data Firehose streams, and deliver to multiple destinations simultaneously. For example, you can route execution logs to Amazon S3 in Apache Parquet format for cost-efficient long-term storage and analysis with Amazon Athena, while simultaneously delivering structured JSON logs to CloudWatch Logs for real-time alerting.

This feature is available in all AWS Regions where API Gateway REST APIs are available, including the AWS GovCloud (US) Regions. Execution logs delivered through this feature are charged at vended logs rates. For pricing details, see Amazon CloudWatch Pricing. You can set up delivery through the API Gateway console, AWS CLI, or AWS CloudFormation. To get started, see Amazon API Gateway documentation and AWS blog post. 

Today, AWS announced that AWS Transform for .NET can automatically generate unit tests for the code it modernizes. When enabled, AWS Transform generates unit tests that target the testable classes in your transformed .NET application, such as business logic and controllers, giving you an automated test safety net on the modernized code as part of the same job that performs the migration.

Modernizing a .NET Framework application to modern .NET produces a transformed, buildable codebase, but teams previously had to write test coverage for that modernized code by hand. With this launch, AWS Transform assesses your application for testability in parallel with the standard .NET assessment, plans which classes and methods to cover, and generates the corresponding unit test code, so you complete the migration with tests already in place. Unit test generation is opt-in: you can enable it at the start of a job or after transformation completes.

Unit test generation for AWS Transform for .NET is supported in the AWS Toolkit for Visual Studio extension. It is available in all AWS Regions where AWS Transform for .NET is supported. To get started, run a .NET transformation with AWS Transform for .NET in Visual Studio and choose to generate unit tests. To learn more, refer to Modernizing .NET in the IDE in the AWS Transform User Guide.

人工知能 (AI) に投資した 1 ドルごとに 2 ドルのリターンが得られるのであれば、コストの増加は非効率で […]

Bulletin ID: 2026-106-AWS
Scope: AWS
Content Type: Important (requires attention)
Publication Date: 09/10/2026 10:00 AM PDT

Description:

Deep Java Library (DJL) is an open-source, engine-agnostic Java framework for deep learning, developed and maintained by Amazon. AWS identified CVE-2026-85228, an integer overflow in the tensor buffer validation component of DJL on all platforms. A crafted tensor payload declaring a shape whose computed byte size exceeds the 32-bit signed integer range causes the size to wrap, allowing an undersized buffer to pass validation; a subsequent tensor operation then reads out of bounds. If leveraged, a remote, unauthenticated actor could obtain information from adjacent process memory or cause a denial of service. A fix for this issue has been released and we recommend users of DJL upgrade to version 0.37.0 or later.

Impacted versions: >=0.13.0 AND <=0.36.0

Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.

Bulletin ID: 2026-107-AWS
Scope: AWS
Content Type: Important (requires attention)
Publication Date: 09/10/2026 11:30 AM PDT

Description:

AWS Systems Manager Agent (SSM Agent) is software that runs on managed nodes (EC2 instances, on-premises servers, and other supported machines) and processes requests from the AWS Systems Manager service, enabling capabilities including Session Manager port forwarding to remote hosts. We identified CVE-2026-89049, a server-side request forgery issue in the remote-host port forwarding functionality. Due to improper validation of equivalent address representations, an authenticated user with port-forwarding permission could bypass the remote destination denylist and reach link-local endpoints, potentially obtaining the managed instance's temporary IAM role credentials and acting with that role's permissions from outside the instance.

Impacted versions: < 3.3.4851.0 (all versions supporting remote-host port forwarding)

Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.

Real-time streaming workers that hold hundreds of persistent WebSocket connections lose data when a worker fails. Learn how to build a WebSocket fleet management system on Amazon ECS and AWS Fargate that uses Amazon DynamoDB conditional writes as a distributed lease to track ownership, fail over automatically, and deploy with low downtime.

Combine zero-shot forecasting with Amazon Chronos2 and multi-agent orchestration on Amazon Bedrock AgentCore to turn demand forecasts into validated purchase orders. No per-product model training, with business rules, auditability, and cost that scales to zero.

AvioBook, a Thales Group Company, prototyped Connected Analytics on Amazon Bedrock AgentCore to turn AvioBook Connect's operational data into plain-language, evidence-based answers for airline managers and dispatchers, helping them find and act on the causes of flight turnaround delays.

Multi-turn agents fail in ways single-turn evaluation misses: one early mistake corrupts every later turn. This post introduces the Agent Evaluation Metric (AEM), a decomposable, turn-level way to measure agent quality, applied to its first dimension, correctness, to pinpoint the turn that caused a failure and separate it from the turns that inherited it.

A configurable, model-agnostic detector that turns any large language model on Amazon Bedrock into a PII detector. Because the entities to detect live in a prompt rather than in code, one detector adapts to new entity types without retraining, and it outperforms an off-the-shelf tool across five public corpora and nine LLM-based detectors.

Learn how to build an end-to-end RFI questionnaire workflow with Amazon Quick Automate. Read a multi-tab RFI workbook from Amazon S3, use natural-language prompts to extract and structure the questionnaire data, refine the workflow through conversation, and write clean CSV output back to Amazon S3 — cutting development from days to hours.

Your teams get an AI assistant that handles real work while your data stays in your environment and your conversations stay private Today, the Amazon Quick desktop application is generally available on macOS and Windows. We’re also adding a new activity feed to the mobile experience on iOS and Android that consolidates email, calendar, CRM, […]

TwelveLabs Marengo Embed 3.0 is now generally available as an embedding model in Amazon Bedrock Knowledge Bases, bringing fully managed natural language search to video, image, and audio content. This walkthrough shows how to build a knowledge base powered by Marengo 3.0 and run semantic queries against your media.

Amazon SageMaker HyperPod now supports model caching for inference, which pre-loads model weights and container images onto cluster nodes so pods read from local NVMe storage instead of downloading over the network. Learn how model caching cuts cold starts from tens of minutes to seconds, how it works, and how to enable it.

Amazon SageMaker Inference now offers prefix-aware routing, a routing strategy that sends requests sharing the same prompt prefix to the same instance so the KV cache stays warm. In benchmarks on Llama 3.1 70B, it reduced P50 time-to-first-token by up to 77% and raised KV cache hit rates from about 25% to over 80%.