コンピューティングと生成 AI の機能強化が幅広く発表された一日でした。NVIDIA RTX PRO 4500 Blackwell GPU を搭載した Amazon EC2 G7 インスタンスが US East (Ohio)・US West (Oregon) で一般提供を開始し、G6 比で最大 4.6 倍の AI 推論性能を実現します。Amazon ECS は 20 秒粒度の高解像度メトリクスにより自動スケーリングを最大 4.2 倍高速化。Amazon Bedrock AgentCore harness と Web Search が一般提供となり、数分で本番グレードのエージェントを構築できるようになりました。SageMaker AI では推論エンドポイントの新しいオブザーバビリティ機能や JumpStart への all-MiniLM-L12-v2・Ministral-3-14B 追加、EKS の顧客経由コントロールプレーン egress、MSK Express の Intelligent Rebalancing 拡大、ベトナム・ハノイの Local Zone 一般提供なども発表。セキュリティ面では Spring 2026 SOC レポートの OSCAL 提供と containerd CRI プラグインの複数 CVE が告知されました。
コンピューティング: EC2 G7 (Blackwell GPU) GA、ECS 高速オートスケーリング、EKS egress ルーティング
生成 AI: Bedrock AgentCore harness / Web Search GA、SageMaker 推論オブザーバビリティと JumpStart 新モデル
メッセージング/HPC: MSK Express Intelligent Rebalancing、PCS の P6e UltraServer 対応
グローバルインフラ: ハノイ Local Zone GA、SNS ソウルリージョン SMS、CloudWatch Synthetics マルチロケーション
セキュリティ: SOC 1/2 レポート OSCAL 提供、containerd CRI プラグイン CVE、Kiro CLI 調査支援
Customers that use Amazon Simple Notification Service (Amazon SNS) in the Asia Pacific (Seoul) Region can now send text messages (SMS) to subscribers in more than 200 countries and territories.
Amazon SNS is a fully managed pub/sub messaging service that enables message delivery to multiple endpoints including AWS Lambda, Amazon SQS, Amazon Data Firehose, mobile devices, and email. With this launch, customers using SNS in the Asia Pacific (Seoul) Region can subscribe phone numbers to SNS topics and broadcast SMS messages via AWS End User Messaging.
To learn more about sending SMS messages with SNS, visit Mobile text messaging with Amazon SNS. For the list of supported countries and regions, visit Supported countries and regions.
Amazon SageMaker AI's new observability capability allows customers to operate production generative AI inference workloads with confidence by providing comprehensive visibility into token performance, GPU health, inference component placement, and autoscaling behavior. It takes away the manual work of searching CloudWatch for per-endpoint metrics, correlating latency spikes with GPU saturation or KV cache exhaustion and diagnosing why scaling operations are slow. This capability tracks inference performance metrics in real-time, including Time to First Token, inter-token latency, queue depth, and tokens per second, and surfaces them alongside infrastructure health so customers can identify and resolve issues in minutes rather than hours.
SageMaker AI detailed observability transforms how customers monitor and optimize their inference fleet. The new pre-built SageMaker AI Insights dashboard in Amazon CloudWatch gives customers token latency, GPU utilization, inference component copy counts, scaling events, and cold start breakdowns in a single view with OpenTelemetry native metrics published automatically, no instrumentation required. This allows teams to quickly diagnose TTFT degradation, verify availability zone compliance, and tune autoscaling policies. Customers who have standardized on observability tools like Grafana can connect directly using the regional PromQL endpoint and import a pre-configured dashboard template. This capability helps customers self-serve operational issues and maximize the performance of their AI investments.
SageMaker AI Inference observability is available in the following AWS Regions: US East (N. Virginia), US East (Ohio), US West (Oregon), US West (N. California), Canada (Central), South America (São Paulo), Europe (Ireland), Europe (Frankfurt), Europe (London), Europe (Stockholm), Europe (Zurich), Asia Pacific (Mumbai), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Tokyo), Asia Pacific (Seoul), and Asia Pacific (Jakarta). To learn more, visit the Documentation and Amazon SageMaker AI webpage.
Today, Amazon Elastic Kubernetes Service (Amazon EKS) introduces customer-routed control plane egress, a capability that lets you route outbound Kubernetes API server traffic through your own Amazon VPC. This includes admission webhook callbacks, OpenID Connect (OIDC) provider lookups, and aggregate API server requests. With customer-routed control plane egress, this traffic flows through your VPC, where you control the routing, security groups, and egress path.
Organizations with data perimeter requirements, compliance mandates, or private network infrastructure can use customer-routed control plane egress to reach private OIDC providers and webhook servers that are accessible only within their VPC, and control how that traffic routes through their network. To get started, set controlPlaneEgressMode to CUSTOMER_ROUTED when creating a new cluster or updating an existing cluster. To enforce this configuration organization-wide, use the eks:controlPlaneEgressMode IAM condition key with AWS Organizations Service Control Policies.
Customer-routed control plane egress is available at no additional cost in all AWS Regions where Amazon EKS is available. To learn more, see Configure control plane egress routing in the Amazon EKS User Guide.
Today, AWS announced the availability of all-MiniLM-L12-v2 in Amazon SageMaker JumpStart, expanding the portfolio of models available to AWS customers. This model from Sentence Transformers maps sentences and paragraphs to a 384-dimensional dense vector space, enabling customers to build high-quality semantic search, text clustering, and sentence similarity applications on AWS infrastructure.
all-MiniLM-L12-v2 excels at encoding sentences and short paragraphs into dense vector representations that capture semantic meaning, making it ideal for information retrieval, semantic search systems, document clustering, duplicate detection, and paraphrase identification. Its compact architecture delivers fast inference while maintaining strong embedding quality, well suited for production workloads that require efficient text representations at scale.
With SageMaker JumpStart, customers can deploy this model with just a few clicks to address their specific AI use cases. To get started with this model, navigate to the Models section of SageMaker Studio or use the SageMaker Python SDK to deploy the model to your AWS account. For more information about deploying and using foundation models in SageMaker JumpStart, see the Amazon SageMaker JumpStart documentation.
Today, AWS announced the availability of Ministral-3-14B-Instruct-2512 in Amazon SageMaker JumpStart, expanding the portfolio of foundation models available to AWS customers. This model from Mistral AI delivers frontier-class multimodal capabilities in a compact 14B-parameter architecture optimized for edge deployment, enabling customers to build advanced AI assistants, agentic systems, and vision-enabled applications on AWS infrastructure.
Ministral-3-14B-Instruct excels at analyzing images and providing insights based on visual content in addition to text, agentic capabilities with native function calling and JSON output, and multilingual understanding across dozens of languages including English, French, Spanish, German, Chinese, Japanese, Korean, and Arabic.
With SageMaker JumpStart, customers can deploy this model with just a few clicks to address their specific AI use cases. To get started with this model, navigate to the Models section of SageMaker Studio or use the SageMaker Python SDK to deploy the model to your AWS account. For more information about deploying and using foundation models in SageMaker JumpStart, see the Amazon SageMaker JumpStart documentation.
AWS Compute Optimizer now includes improved visibility into IOPS and throughput spikes when deliverings Amazon EBS volume rightsizing recommendations. Compute Optimizer analyzes two additional Amazon CloudWatch metrics, VolumeIOPSExceededCheck and VolumeThroughputExceededCheck, which report whether your workload consistently attempted to drive IOPS or throughput beyond your volume's provisioned performance in any given minute. By factoring in these signals, Compute Optimizer helps you make rightsizing decisions to balance cost with performance for workloads that experience bursts of high IOPS or throughput.
This enhancement is available in all AWS Regions where AWS Compute Optimizer is available, except the AWS GovCloud (US) Regions, and the China Regions. The underlying CloudWatch metrics are available at no additional charge for all EBS volumes attached to Nitro-based EC2 instances, excluding standard and Multi-Attach enabled volumes. To get started, go to AWS Compute Optimizer in the AWS Management Console. To learn more, visit the AWS Compute Optimizer User Guide.
Amazon Connect Customer now supports the ability to interrupt an agent with a contact, overriding their usual routing configuration in case of urgent or time-sensitive work. For example, an agent may be waiting for a time-sensitive callback on their personal extension, while taking customer service calls in the meantime. When that urgent call comes in, it can now ring the agent even if the agent is currently already on another call, so the agent can decide whether to put the first caller on hold to pick up the callback as well.
You can also use this feature to directly assign certain contacts to a specific agent even though that agent has set themselves to a custom status where they normally could not be offered queued contacts. For example, you may want to ensure that a specific agent cannot take customer service calls while in “Back Office Work” but still allow calls to their personal extension to ring through, improving efficiency for urgent contacts.
This feature is available in all AWS regions where Amazon Connect Customer is offered. To learn more about this feature, see the Amazon Connect Customer Administrator Guide. To learn more about Amazon Connect Customer, the AWS cloud-based contact center, please visit the Amazon Connect Customer website.
Starting today, Nested virtualization is now available on additional Intel platforms and additional Regions. Nested virtualization is now available on C7i,R7i, M7i, C7id,R7id, M7id, C7i-flex,R7i-flex, M7i-flex, I7i, C8i-flex,R8i-flex, M8i-flex,and X8i, in addition to already available support on C8i, M8i and R8i instances. This capability is also now available in US GovCloud (US-East) and US GovCloud (US-West), in addition to existing support in all commercial regions.
With nested virtualization capabilities, customers can create nested environments by running KVM or Hyper-V on virtual EC2 instances. Customers can leverage this capability for use cases such as running emulators for mobile applications, simulating in-vehicle hardware for automobiles, and running Windows Subsystem for Linux on Windows workstations. To learn more see documentation .
Starting today, Nested virtualization is now available on additional Intel platforms and additional Regions. Nested virtualization is now available on C7i,R7i, M7i, C8id,R8id, M8id, C7i-flex, M7i-flex, I7i, C8i-flex,R8i-flex, M8i-flex,and X8i, in addition to already available support on C8i, M8i and R8i instances. This capability is also now available in AWSGovCloud (US-East) and AWS GovCloud (US-West), in addition to existing support in all commercial regions.
With nested virtualization capabilities, customers can create nested environments by running KVM or Hyper-V on virtual EC2 instances. Customers can leverage this capability for use cases such as running emulators for mobile applications, simulating in-vehicle hardware for automobiles, and running Windows Subsystem for Linux on Windows workstations. To learn more see documentation .
AWS Parallel Computing Service (PCS) now supports Amazon EC2 P6e-GB200 and P6e-GB300 UltraServer instances, enabling customers to run large-scale GPU workloads using the NVIDIA Blackwell architecture within Slurm-managed clusters. You can reserve UltraServers through EC2 Capacity Blocks for ML, associate them with a PCS compute node group via an EC2 launch template, and PCS automatically configures Slurm with the correct topology plugin.
With P6e-GB200 UltraServers, you can access up to 72 NVIDIA Blackwell GPUs within one NVLink domain to use 360 petaflops of FP8 compute (without sparsity) and 13.4 TB of total high bandwidth memory (HBM3e). P6e-GB300 UltraServers provide 1.5x GPU memory and 1.5x FP4 compute (without sparsity) compared to P6e-GB200.
AWS PCS is a managed service that simplifies running and scaling HPC workloads on AWS using Slurm. You can build complete, elastic environments that integrate compute, storage, networking, and visualization tools, while the service handles cluster operations with managed updates and built-in observability features.
You can use P6e UltraServers with PCS in all AWS Regions where both PCS and EC2 Capacity Blocks for UltraServers are available. To learn more about P6e UltraServers, visit Amazon EC2 P6 instances. To reserve P6e UltraServers, contact your AWS sales representative. Read more about PCS support for P6e UltraServers in the PCS User Guide and make sure to set the right Permissions.
Amazon MQ for RabbitMQ now supports private networking, enabling your brokers to connect to private resources in your VPC without exposing those resources publicly.. This helps you meet your security and compliance requirements when your brokers need to reach private identity providers (such as LDAP and OAuth 2.0), other Amazon MQ for RabbitMQ brokers, or self-hosted RabbitMQ brokers. Previously, this connectivity for RabbitMQ Federation, Shovel, or authentication required Network Load Balancer and NAT Gateway workarounds.
Amazon MQ establishes this connectivity using Amazon VPC Lattice, AWS Resource Access Manager (AWS RAM), and AWS PrivateLink, and manages the underlying infrastructure on your behalf. To get started, create a VPC Lattice resource gateway, package your resource configurations into an AWS RAM resource share, and associate it with your broker.
Private networking is available only for Amazon MQ for RabbitMQ brokers, in all AWS Regions where Amazon VPC Lattice is available. To learn more, see Private networking in the Amazon MQ Developer Guide and the Amazon MQ pricing page.
Today, AWS announces the general availability of Amazon Elastic Compute Cloud (Amazon EC2) G7 instances, accelerated by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. G7 instances deliver up to 4.6x AI inference performance and up to 2.1x graphics performance compared to G6.
You can use G7 instances for AI inference workloads such as language translation, video and image analysis, speech recognition, and recommender systems. Additionally, G7 instances also accelerate graphics workloads such as creating and rendering real-time, cinematic-quality graphics, and game streaming, as well as data analytics workloads such as large-scale data processing pipelines. G7 instances feature up to 8 NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs with 32 GB of memory per GPU, custom Intel Xeon 6 processors, and up to 700 Gbps of Elastic Fabric Adapter (EFA) networking bandwidth.
You can start using Amazon EC2 G7 instances today in two AWS Regions: US East (Ohio) and US West (Oregon). You can purchase G7 instances as On-Demand Instances, as part of Savings Plans, or Spot Instances.
To get started, visit the AWS Management Console, AWS Command Line Interface (CLI), and AWS SDKs. To learn more, visit this blog post and the G7 instance page.
Amazon ECS service auto scaling now detects and responds to load changes faster with support for high resolution (20-second) metrics and metric publishing optimizations. In AWS benchmarking tests, time to trigger scale-out improved from 363 seconds to 86 seconds (76% faster, 4.2x), and total time to scale and provision new tasks improved from 386 seconds to 109 seconds (72% faster, 3.5x). Faster service auto scaling also enables you to reduce baseline capacity and lower compute costs while maintaining service reliability and performance as workload demand fluctuates.
Amazon ECS service auto scaling automatically adjusts task counts to meet workload demand with comprehensive scaling policies, including predictive scaling for recurring traffic patterns, scheduled scaling for planned events, and target tracking to scale dynamically on real-time metrics. With today's launch, target tracking policies for CPU and memory utilization now support 20-second metric resolution, in addition to the default 60-second resolution, for faster scaling signal detection. To get started, use the AWS Console, CLI, CloudFormation, or AWS SDKs to configure 20-second resolution for CPU or memory utilization metrics when creating or updating your ECS service, then configure a target tracking policy selecting the corresponding high-resolution predefined metric.
This feature is available in all AWS commercial and AWS GovCloud (US) Regions, across all ECS compute options: AWS Fargate, Amazon ECS Managed Instances, and Amazon EC2. High-resolution metrics are subject to standard CloudWatch charges; for a pricing example, see Amazon CloudWatch pricing. To learn more, see our documentation and the launch blog post.
Amazon MSK Provisioned clusters with Express brokers now support Intelligent Rebalancing on all existing clusters, at no additional cost. Previously available only on newly created clusters, Intelligent Rebalancing is now available on all MSK Provisioned clusters running Express brokers, making it effortless for customers to benefit from automatic partition balancing when scaling their Express-based clusters up or down.
Intelligent Rebalancing maximizes the capacity utilization of MSK Express-based clusters by optimally rebalancing Kafka resources for better performance, eliminating the need for customers to manage partitions themselves or via third-party tools. Intelligent Rebalancing performs these operations up to 180 times faster compared to Standard brokers. Clusters are continuously monitored for resource imbalance or overload based on intelligent Amazon MSK defaults to maximize cluster performance. When required, brokers are efficiently scaled without affecting cluster availability for clients to produce and consume data.
Intelligent Rebalancing is now available on all MSK Provisioned clusters with Express brokers in all AWS Regions where Express brokers are available. To learn more, see the Amazon MSK Developer Guide.
Today, Amazon CloudWatch Synthetics announces support for multilocation canaries, allowing developers and site reliability engineers to run the same canary across multiple AWS Regions simultaneously from a single point of management. Previously, monitoring application availability from multiple geographic locations required creating and managing separate canaries in each Region, adding operational overhead and increasing the risk of configuration drift. With multilocation canaries, you create and manage a canary in one primary Region, and CloudWatch Synthetics automatically replicates it to the additional Regions you choose, consolidating all run data, metrics, and artifacts in the primary Region.
Multilocation canaries help you ensure consistent user experience worldwide, identify region-specific performance bottlenecks, and validate that third-party dependencies like CDNs and payment processors work across all locations. Replica canaries run independently, giving you resilient monitoring coverage across geographic locations. You can also configure alarms that activate only when issues are detected from multiple locations, increasing alert confidence and helping your team focus on real customer-impacting problems. Amazon CloudWatch Synthetics multilocation canaries are available in all AWS commercial Regions that support CloudWatch Synthetics. You can upgrade existing single-region canaries to multilocation by adding replica Regions without recreating them. For more information about regional availability, see the AWS Region table.
To learn more about CloudWatch Synthetics, see Using synthetic monitoring in the Amazon CloudWatch User Guide. To get started, visit the Amazon CloudWatch product page.
Today, AWS announces the general availability of a new Local Zone in Hanoi, Vietnam, bringing AWS infrastructure closer to end users. This new Local Zone is one of the first AWS Local Zones in the Asia Pacific with support for Amazon Simple Storage Service (Amazon S3) and Amazon Elastic Block Store (Amazon EBS) Local Snapshots, enabling customers to meet data residency requirements by storing and backing up data locally.
AWS Local Zones are AWS infrastructure deployments that extend core services, such as compute, storage, networking, and other select services, closer to metropolitan areas worldwide. AWS Local Zones help you achieve single-digit millisecond latency for end-user workloads, meet data residency requirements, support AI/ML inference workloads, and accelerate migration and modernization of legacy applications to the cloud, all while maintaining consistent AWS APIs, tools, and services as AWS Regions. AWS Local Zones are available in more than 30 metropolitan areas worldwide.
The Hanoi Local Zone supports Amazon Elastic Compute Cloud (Amazon EC2) with C7i, M7i, and R7i instances, Amazon S3 with the One Zone-Infrequent Access storage class, Amazon EBS with Local Snapshots and volume types gp3, gp2, io1, sc1, and st1, Amazon Elastic Container Service (Amazon ECS), Amazon Elastic Kubernetes Service (Amazon EKS), Amazon Virtual Private Cloud (Amazon VPC), AWS Direct Connect, and Application Load Balancer.
To get started, enable the Hanoi Local Zone (ap-southeast-1-han-1a) from the Regions and Zones tab in the AWS Global View or by using the ModifyAvailabilityZoneGroup API. For pricing information, visit the AWS Local Zones pricing page. To learn more, visit the AWS Local Zones overview page.
Amazon Elastic Container Service (Amazon ECS) service auto scaling automatically adjusts task counts to meet workload demand with comprehensive scaling policies, including predictive scaling for recurring traffic patterns, scheduled scaling for planned events, and target tracking to scale dynamically on real-time metrics. You can choose proactive scaling by using predictive scaling (automatic) and scheduled scaling […]
Announcing the general availability of Amazon Elastic Compute Cloud (Amazon EC2) G7 instances, delivering high performance GPU acceleration for AI inference, graphics, and data analytics workloads.
はじめに 通信業界の主要 MNO 4 社（NTTドコモ、KDDI、ソフトバンク、楽天モバイル）に NTT、ソニ […]
みなさんこんにちは！ 猫や犬に癒されるより AI bou (相棒) に助けられる日々を送る Solution […]
このテレオペのデモでは AWSのIoTサービスを利用し、 WebのUIを見ながらゲームコントローラーを操作することで、クラウド経由でロボットを操作します。
本ブログは、コニカミノルタ株式会社と Amazon Web Services Japan が共同で執筆しました […]
Amazon Web Services (AWS) is excited to release the Spring 2026 System and Organization Controls (SOC) 1 and 2 reports in machine-readable OSCAL format alongside the PDF version of the reports. The reports cover 188 services over the 12-month period from April 1, 2025 to March 31, 2026, giving customers a full year of assurance. […]
When a security event occurs in your Amazon Web Services (AWS) environment, rapid response is critical. However security teams often struggle with time-consuming, manual processes that slow down investigations. Analysts must recall complex AWS Command Line Interface (AWS CLI) syntax for multiple services, manually correlate findings across Amazon GuardDuty, AWS CloudTrail, and other security tools, […]
Bulletin ID: 2026-046-AWS
Scope: AWS
Content Type: Important (requires attention)
Publication Date: 06/18/2026 17:30 PM PDT
Description:
containerd is an open-source container runtime used by Kubernetes via the Container Runtime Interface (CRI) plugin. It underpins AWS managed container services including Amazon Elastic Kubernetes Service (Amazon EKS), Amazon Elastic Container Service (Amazon ECS), AWS Fargate, Bottlerocket, and Amazon Linux. AWS identified five issues in the containerd CRI plugin affecting versions 1.7 through 2.3.
- CVE-2026-50195 (GHSA-cvxm-645q-p574) - CRI checkpoint import, local image tag poisoning
- CVE-2026-53488 (GHSA-xhf5-7wjv-pqxp) - image-config LABEL -> host-root command exec
- CVE-2026-53492 (GHSA-33vj-92qq-66hc) - CDI annotation smuggling during checkpoint restore
- CVE-2026-53489 (GHSA-rgh6-rfwx-v388) - arbitrary host file read via symlink in checkpoint restore
- CVE-2026-47262 (GHSA-jpcc-p29g-p8mq) - image-triggered runtime DoS
Impacted versions: containerd 1.7, 2.0, 2.1, 2.2, 2.3
Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.
Today, Amazon Bedrock AgentCore harness is generally available. Two API calls (CreateHarness to define an agent, and InvokeHarness to run it), and you have an agent running in seconds. The agent runs in its own isolated environment with a filesystem and shell, so it can read files, run commands, and write code safely. It remembers users and conversations across sessions, picks up skills you point it at (including the AWS-curated catalog), browses the web, calls your tools through gateway or MCP, and switches model providers mid-session without losing context. Every step streams back to you in real time and is automatically traced to Amazon CloudWatch. You don’t need to write orchestration code or build a container, unless you want to.
Amazon SageMaker AI provides fully managed real-time inference hosting for machine learning models. You deploy a model to a SageMaker endpoint backed by one or more compute instances, and SageMaker handles provisioning and scaling. SageMaker supports multiple endpoint architectures. This post focuses on the two most relevant to generative AI workloads with detailed observability: Single-model endpoints (SME) and Inference component (IC) endpoints.
This post shows how to enable Adobe Marketing Agent for Amazon Quick using a Model Context Protocol (MCP). We walk you through how to configure the integration, authenticate using your Adobe credentials, and get the latest insights in Amazon Quick. The sample workflow returns audience rankings, loyalty segment summaries, journey usage, and conflict recommendations.
Web Search on Amazon Bedrock AgentCore is now generally available. In this post, we walk through what makes Web Search on Amazon Bedrock AgentCore different, why it matters, and how to wire it in with a few lines of code.