AWS News - 2026-09-30

We're excited to announce the availability of Anthropic's Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5 on Amazon Bedrock in India, Claude Opus 5 and Claude Sonnet 5 in South Korea, as well as Claude Sonnet 5 in Singapore. Customers who need to process data within a specific geography, including those in financial services, healthcare, and the public sector, can now use these models at scale while keeping inference in-country.

In India, the regional endpoint is served through geographic cross-Region inference across the Mumbai and Hyderabad Regions. Customers in India can access these models while processing data in India Regions, in addition to the already supported global cross-Region inference.

In South Korea, Amazon Bedrock supports Claude Opus 5 and Claude Sonnet 5 models with in-region inference on the bedrock-runtime endpoint in the Seoul (ap-northeast-2) Region. In Singapore (ap-southeast-1), Amazon Bedrock supports Claude Sonnet 5 with in-region inference on the bedrock-runtime endpoint. Amazon Bedrock processes inference requests and data within the Region you call, and the processing never leaves that Region.

For the most current information about model availability in each Region, see Regional availability by models in the Amazon Bedrock User Guide.

Amazon Relational Database Service (RDS) for PostgreSQL announces Amazon RDS Extended Support minor versions 13.23-rds.20260514, 12.22-rds.20260514, and 11.22-rds.20260514.

Amazon RDS Extended Support provides up to three additional years of fixes for critical Common Vulnerabilities and Exposures (CVE) and bugs beyond a major version's end of standard support date, giving you more time to upgrade to a new major version.

You can upgrade your database using Amazon RDS Blue/Green Deployments, in-place upgrade, or restore from a snapshot. Learn more about performing version upgrades in the Amazon RDS for PostgreSQL User Guide. 

Amazon RDS for PostgreSQL makes it simple to set up, operate, and scale PostgreSQL deployments in the cloud. Create or update a fully managed Amazon RDS database in the Amazon RDS Management Console or by using the AWS Command Line Interface (AWS CLI).

Amazon Aurora PostgreSQL-Compatible Edition now supports PostgreSQL versions 18.6, 17.11, 16.15, 15.19, and 14.24, which include bug fixes from the PostgreSQL community and Aurora-specific enhancements. We recommend upgrading to the latest minor versions to benefit from these improvements and address Common Vulnerabilities and Exposures (CVEs), as detailed in the release notes.

You can upgrade your databases during scheduled maintenance windows using automatic minor version upgrades. To simplify operations at scale, enable automatic minor version upgrades and use the AWS Organizations Upgrade Rollout Policy to orchestrate multiple upgrades in phases, validating on lower-priority environments before upgrading your most critical ones. For more information, see Upgrading Amazon Aurora PostgreSQL DB clusters.

Amazon Aurora is designed for high performance and availability at global scale with full PostgreSQL compatibility. It provides scale-to-zero serverless compute, Aurora Global Database for multi-Region resilience, Aurora I/O-Optimized for improved price performance on I/O-intensive workloads, and built-in security and continuous backups. To get started, take a look at our getting started page.

Amazon Connect Customer now lets administrators store more of the reference data that drives their contact center configurations. Administrators can now use data tables to hold up to 20,000 values per table for defining configurations such as operating hours, escalation rules, AI agent prompts or queue assignments that drive contact flows. Administrators can now also export tables, edit hundreds of records, move tables between Connect Customer instances, or load reference data from external systems using tools and sources such as Excel and DynamoDB. Business teams can keep this data current themselves, adjusting their contact center configurations as conditions change without depending on technical resources.

This feature is available in all AWS commercial and AWS GovCloud (US-West) regions where Amazon Connect Customer is offered. To learn more, see Create and configure data tables in the Amazon Connect Administrator Guide. To learn more about Amazon Connect, visit the Amazon Connect Customer website .

Today, Amazon WorkSpaces Applications announces unified graphics images, a single image type that works across all supported graphics instance families, including G4dn, G5, G6, and G7. Previously, each graphics generation required its own dedicated image. Now customers can use one image to create and manage fleets on any supported graphics instance family.

Unified graphics images simplify image management and make it easier to adopt newer GPU generations as they become available. When creating a fleet using the unified graphics image, the console displays all compatible instance types, so you can choose the right price and performance for your workload. Fleets built on a unified graphics image can also change instance types across supported families using the console or the UpdateFleet API, an operation previously limited to sizes within a single instance family.

Unified graphics images are available in all AWS commercial and AWS GovCloud (US) Regions where Amazon WorkSpaces Applications graphics instances are supported.

To get started, upgrade an existing graphics image through Managed Image Update or launch an image builder from a unified base image. To learn more, visit the Amazon WorkSpaces Applications documentation.

Today, AWS announces the general availability of GPT-6.1 Sol from OpenAI on Amazon Bedrock. An upgrade to GPT-6 Sol, GPT-6.1 Sol delivers exceptionally strong performance on agentic coding, computer use, and professional work. According to OpenAI, it approaches GPT-6 Astra across demanding evaluations at roughly one-fifth of the cost, giving developers more room to build and run capable agents at scale. The Amazon Bedrock inference engine delivers the performance, security, and reliability required for production workloads.

Use GPT-6.1 Sol to build features, debug issues, iterate on solutions, carry out multistep workflows, and investigate complex codebases. It also supports explicit prompt caching on Amazon Bedrock, making it practical for agentic workloads that reuse context across repeated requests. Established AWS controls help you secure workloads, govern access, and audit model invocation activity.

You can get started in the Amazon Bedrock console or programmatically through supported Amazon Bedrock APIs. For information about supported AWS Regions, endpoints, APIs, features, inference profiles and pricing, see the Amazon Bedrock documentation. To learn more, read the blog. 

AWS Deadline Cloud now supports running jobs using Docker containers from Amazon Elastic Container Service (ECS) on Linux-based service-managed fleets. Deadline Cloud is a fully managed service that helps teams run compute-intensive workloads in the cloud for visual effects, animation, product design, simulation, and gaming.

Deadline Cloud customers can now bring their own software containers directly to the service using Docker-enabled fleets. Customers that already package software into container images can point Deadline Cloud to these images directly. This unlocks any container-deployed software use case, from rendering tools like Blender to physical AI workloads such as NVIDIA Isaac Sim and MuJoCo. With service-managed fleets, Deadline Cloud automatically deploys the containerized software on every worker instance at scale.

Containers are integrated with Deadline Cloud fleets and queues, so that customer job templates do not need to contain container commands or image references. This means the same job bundles customers run locally with the Open Job Description CLI can be submitted to Deadline Cloud with no additional changes. The result is a faster path from existing containerized software workflows to a fully managed render and compute farm.

Container support for service-managed fleets is available in all AWS Regions where Deadline Cloud is offered.

To get started, see our user guide and our developer guide.

Amazon RDS now displays the full snapshot size for RDS Snapshots. With this enhancement, customers can now retrieve full snapshot sizes programmatically through the DescribeDBSnapshots API using the new field, ‘FullSnapshotSizeInBytes’. They can also view this on the console with the ‘Full Snapshot Size’ column.   

Amazon RDS snapshots are incremental. This means that if you take multiple snapshots of a volume over time, each snapshot only stores the new or modified blocks while maintaining references to unchanged blocks from previous snapshots. The ‘FullSnapshotSizeInBytes’ field shows you the total size of all blocks that make up a snapshot, including both the blocks stored directly in that snapshot and all blocks referenced from previous snapshots. For instance, if you have a 100 GB database volume with 50 GB of data, the ‘full snapshot size’ would show 50 GB regardless of whether it's the first snapshot or a subsequent one. Please note that this is different from the incremental snapshot size, which only refers to the size of newly changed blocks stored in that specific snapshot.

Full snapshot size is available all Amazon RDS database instances in all commercial AWS Regions. You can start using it today through the Amazon RDS Management Console, the AWS Command Line Interface (CLI), or the AWS SDKs. To learn more, see the Amazon RDS User Guide.

Developed jointly by AWS and OpenAI, Bedrock Managed Agents (BMA) is built on a customized version of OpenAI's Agents API engineered to be AWS-native and integrated with AWS resources. You can now build agents optimized for OpenAI models that run entirely inside AWS with the identities, permissions, and governance controls you already use. 

BMA manages how the model preserves state, selects and uses tools, executes code, and coordinates work across multiple steps and decisions. Durable sessions retain messages, tool calls, and intermediate results so you can return later, provide new information, and continue from the progress already made. You can add reusable skills for specialized procedures and connect agents to tools, including via Model Context Protocol (MCP) servers. Each agent operates with its own IAM role, supports human approval before consequential actions, and records supported API activity with AWS CloudTrail.

During preview, there is no additional charge for BMA beyond the underlying AWS resources your agents consume. Pricing is subject to change at general availability.

Get started with Amazon Bedrock Managed Agents today through supported APIs. BMA is available in preview in US East (N. Virginia), US West (Oregon), and US East (Ohio). For details, see the Amazon Bedrock documentation.  

AWS Deadline Cloud now supports flexible expressions and rich parameter types in job templates, giving customers a cleaner, more powerful way to fit the service to existing production pipelines. Deadline Cloud is a fully managed service that helps teams run compute-intensive workloads in the cloud for visual effects, animation, product design, simulation, and gaming.

Customers can now express pipeline logic directly in their job templates. This includes expressions such as arithmetic, conditionals, string and path operations, and list comprehensions, all in a familiar Python-style syntax. Customers can use this to split a frame range into per-task frames, derive an output path from the scene name, or toggle an optional flag directly within job templates themselves, making it easier to automate bespoke pipeline operations. New boolean, list, and frame-range parameter types match how pipelines already describe work. For example, a checkbox can add a --gpu flag and its matching host requirement when set, and a list of cameras can become a step's task range. Expressions are type-checked at submission, and their evaluation is deterministic, meaning that job execution behaves predictably every time.

These capabilities come from the new EXPR extension to Open Job Description, the open specification Deadline Cloud uses for job templates, and are available in all AWS Regions where AWS Deadline Cloud is supported.

To get started, visit the Deadline Cloud developer guide.

Amazon S3 Tables add support for geometry, geography, unknown, and nanosecond timestamp data types, along with column default values, as defined in the Apache Iceberg Version 3 (V3) specification. You can now store geospatial coordinates and nanosecond-precision event times natively instead of encoding them in strings or integers. This helps simplify data architecture while improving storage efficiency and performance. With this launch, S3 Tables support all data types introduced in V3, adding to existing support for the Variant data type, deletion vectors, and row lineage.

Geometry and geography columns store points, lines, and polygons natively, so fleet tracking and asset mapping workloads can filter on location at query time. Nanosecond timestamps let telemetry and financial workloads record event times at source precision. Column default values populate a newly added column for existing rows, with no backfill required. S3 Tables provide automatic table maintenance and compaction for Apache Iceberg tables, so tables using V3 data types stay performant and cost effective as data scales.

Support for these data types in S3 Tables is available in all AWS Regions where S3 Tables are available. To learn more, see Amazon S3 Tables, Apache Iceberg V3 on AWS, and the AWS Prescriptive Guidance for working with Iceberg table format specification version 3.

AWS announces the general availability of the AWS Marketplace metering agent skill, an AI-guided experience that helps sellers build, deploy, and validate a usage-based (pay-as-you-go) SaaS metering integration from within their AI coding assistant. The skill guides sellers through the complete integration journey — gathering their product type, pricing model, and usage dimensions; recommending the correct metering API; generating integration code and a customized AWS CloudFormation stack tailored to their configuration; and running an end-to-end test against AWS Marketplace before any production code ships.

Previously, sellers integrating metering navigated documentation, workshops, and trial-and-error API calls, where common mistakes such as mismatched dimension names or invalid timestamps create billing gaps discovered days later. The metering agent skill cross-validates dimensions against the seller's actual product configuration, applies built-in guardrails, deploys a serverless metering pipeline (using ResolveCustomer API, BatchMeterUsage API, and Amazon EventBridge for subscription events), and verifies the integration with a live test. It also supports Concurrent Agreements and helps existing sellers inspect, debug, and analyze their metering records.

The skill is available through the AWS MCP Server in any AI coding assistant that supports MCP, including Amazon Q Developer, Kiro, and other MCP-compatible clients, with no plugin installation required.

To get started, visit Configuring metering for usage with SaaS subscriptions.

AWS Transfer Family now automatically approves requests to increase your SFTP connector quota up to 1,000 connectors per AWS account in each AWS Region. You can add more connectors as your file transfer needs grow without waiting for manual approval.

The default quota remains 100 SFTP connectors per AWS account in each AWS Region, and you continue to request increases in the AWS Service Quotas console. Previously, all quota increases required manual approval. Automatic approval applies to connectors using either service-managed or Amazon VPC Lattice egress. If you need more than 1,000 connectors, submit a higher quota request in the Service Quotas console.

This feature is available in all AWS Regions where Transfer Family SFTP connectors are offered. To learn more, visit the AWS Transfer Family SFTP connectors documentation.

AWS Accounts now support phone number verification for primary contact phone numbers. Previously, phone numbers in AWS account contact information were validated for format but never verified through an out-of-band mechanism. Now, customers can verify their phone numbers via SMS one-time passcode (OTP).

To verify a phone number, customers initiate verification through the AWS Management Console or programmatically via the new SendPhoneNumberVerification API, which sends a 6-digit OTP. After entering the code, the VerifyPhoneNumber API validates and persists the verified status. When a phone number is changed through the PutContactInformation API, customers will be prompted to complete verification again. The GetContactInformation API now exposes verification status, enabling customers to confirm which phone numbers have been verified.

For AWS Organizations customers, phone number verification supports inheritance from the management account. When a verified phone number is applied to member accounts from the management account, those member accounts inherit the verified status if the number matches the management account, eliminating the need to verify the same number across thousands of accounts. Member accounts that independently change their own phone numbers will need to complete verification separately.

This feature is available in all AWS commercial regions. To learn more about phone number verification, visit the AWS Account Management documentation.

AWS IAM Identity Center's Identity Store APIs now accept the Amazon Resource Name (ARN) for a user, group, group membership, or identity store anywhere the APIs previously accepted the resource ID. ARN support is additive and existing integrations continue to work unchanged.

If you build on the Identity Store APIs, you may already hold resource ARNs — for example, from IAM policy evaluation, CloudTrail events, or cross-service integrations. Previously, you had to strip the ARN down to the resource ID before calling Identity Store APIs. With this change, you can pass either form directly, simplifying application code and eliminating potential parsing errors.

The change applies to every request identifier field across the Identity Store API surface. Responses continue to return resource IDs as they did before. Malformed or wrong-resource-type ARNs return a ValidationException.

This capability is available in all AWS Regions where AWS IAM Identity Center is offered, at no additional cost. To learn more, see the Identity Store API Reference.

Amazon Kinesis Video Streams now supports interface VPC endpoints powered by AWS PrivateLink, providing private connectivity from your Amazon Virtual Private Cloud (Amazon VPC). Traffic between your VPC and Kinesis Video Streams stays on the AWS network and is not exposed to the public internet, across the control plane and the video ingestion and playback data planes.

Customers with strict security, compliance, or network-isolation requirements can now ingest, store, and play back video without internet gateways, NAT devices, or public IP addresses. For example, connected-camera or IoT video workloads in a private subnet can send media to Kinesis Video Streams and retrieve it for playback and analytics entirely over private connectivity. You create the endpoint from the Amazon VPC console, AWS CLI, or AWS SDKs and can attach a VPC endpoint policy to control access.

VPC endpoints for Amazon Kinesis Video Streams are available in all AWS Regions where Amazon Kinesis Video Streams is available, including the AWS GovCloud (US) Regions and the China (Beijing) Region, operated by Beijing Sinnet Technology Co., Ltd. ("Sinnet").

To learn more, see our Getting Started Guide.

Partner Revenue Measurement (PRM) provides measurement of AWS consumption driven by Partner solutions. Today, PRM adds multi-Partner support to Resource Tagging, so that multiple AWS Partners can receive revenue attribution for the same AWS resource. Previously, a resource carried only one Partner tag, aws-apn-id, so when multiple Partners contributed to the same workload, only one Partner received attribution. 

With this update, each Partner tags the resources they contribute to using a new key in the format aws-apn-id-<partner-central-aws-account-id>. When more than one Partner tags the same resource, each tagged Partner receives revenue attribution. This benefits Partners that co-deliver solutions, such as a consulting Partner that deploys and operates a software Partner's product in an AWS account. The existing aws-apn-id key continues to work, and resources tagged with it require no additional changes.  

Multi-Partner support in Resource Tagging is available today in all commercial AWS regions. Revenue attributed through the existing aws-apn-id key and the new key both appear under the “Resource Tagging” method in the Attributed Revenue Dashboard, with no changes to the dashboard.

To learn more, see the Partner Revenue Measurement product page and the Resource Tagging onboarding guide.

長期にわたって保守される組み込みソフトウェアの開発でも、AI エージェントは活用できます。2026 年 7 月、日立産業制御ソリューションズ様向けに、既存コードの読み解きから仕様化、実装、テストまでを Kiro で体験するハンズオンワークショップを 2 回開催し、102 名のエンジニアにご参加いただいた開催報告です。

規制産業では、AI 応答が定められたポリシーに準拠していることを、数学的な確実性をもって示すことが求められます。2 部構成の第 1 部となる本記事では、形式的検証の技術によって生成 AI の出力を検証する Amazon Bedrock Guardrails の自動推論チェックについて、その技術的な基礎を解説します。7 種類の検出結果タイプ、シナリオ生成やテスト管理などの新機能に加え、病院再入院リスク評価を例に、自然言語のドキュメントからのポリシー作成、テストによる検証、注釈を用いたポリシー改善、ガードレールへの統合までを実装手順とともに紹介します。

Bulletin ID: 2026-119-AWS
Scope: AWS
Content Type: Important (requires attention)
Publication Date: 09/29/2026 08:00 AM PDT

Description:

GluonTS is an open source library for deep learning based time series models. We identified CVE-2026-100308 that allows arbitrary command execution upon deserialization of untrusted model artifacts.

Deserialization of untrusted data in the model loading component in Amazon GluonTS before 0.17.0 might allow context-dependent attackers to execute arbitrary operating system commands with the privileges of the loading process via a crafted serialized model directory. The issue arises whenever a user calls Predictor.deserialize() or any RepresentablePredictor.deserialize() on a model directory they do not fully control.

Impacted versions: <0.17.0

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

View article

Condé Nast's editorial teams spent an average of 250 minutes per task searching a library of more than 140,000 videos using only titles and descriptions. Working with the AWS Generative AI Innovation Center, they built a multimodal video discovery solution on Amazon Bedrock and Amazon OpenSearch Service that cut discovery time to under 2 minutes.

Manually extracting data from hundreds of vendor contracts doesn't scale, and RAG chat tools fall short on portfolio-wide questions. This post shares a contract intelligence platform on AWS that uses AI agents to extract and verify contract fields, then answers aggregate and single-contract questions through Amazon Quick analytics.

Part 2 of our Amazon Quick prompt engineering series goes component by component. Learn the prompt patterns that get the best results from Amazon Quick Research, Quick Flows, Quick Sight, chat agents, and action integrations, plus the common pitfalls to avoid.

Prompt engineering in Amazon Quick shapes how accurately its AI-powered features respond to your requests. Part 1 of a two-part series covers the foundational principles and reusable frameworks (specificity, context-setting, few-shot examples, and the CRISPE framework) for consistent, high-quality results across Amazon Quick.

GPT-6.1 Sol is now generally available on Amazon Bedrock, bringing stronger reasoning to coding, computer use, and professional workloads that run frequently.

Amazon Bedrock now supports Anthropic's Claude Opus 5 and Claude Sonnet 5 with in-region inference in Seoul, and Claude Sonnet 5 in Singapore. If you have local data processing requirements in South Korea or Singapore, you can now use these Anthropic models at scale, with inference processed entirely within the Region you call.

Anthropic's Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5 are now available in India through Amazon Bedrock geographic cross-Region inference. You can access these models while processing data within the India Regions, and get started from the Amazon Bedrock console or with the Messages, InvokeModel, and Converse APIs.

Learn a serverless, two-account pattern for running regulated, air-gapped RHEL-family fleets on AWS. It separates connected package ingestion from the air-gapped workload, adds explicit human approval for package changes, and keeps EC2 Image Builder AMIs and Patch Manager on one frozen Amazon S3 repository snapshot.

Managing Amazon EC2 AMIs at scale means constantly mapping AMI IDs to their AWS Systems Manager parameter paths by hand. A recent enhancement to the DescribeImages API returns the associated SSM parameter directly. This post shows how to use it across the AWS CLI, AWS CloudFormation, Terraform, and Auto Scaling launch templates.

Lambda durable functions checkpoint execution state to durable storage, and for regulated payment workloads that data is sensitive. This post shows how to configure a customer managed key in AWS KMS to encrypt durable execution data, define a least-privilege key policy, and verify encryption through AWS CloudTrail, all deployed with Terraform.