この日は生成 AI とエージェント関連のアップデートが目立ちました。Amazon Bedrock AgentCore では Web Search が一般提供となり、AI エージェントがデータレジデンシーを保ったまま最新のウェブ知識を参照できるようになりました。Amazon SageMaker AI は NVIDIA Nemotron 3 Nano モデルのサーバーレスファインチューニング (SFT/RFT) に対応しました。インフラ面では、ストレージ最適化された Amazon EC2 I7i インスタンスが欧州 (パリ) で提供開始となり、Amazon Lightsail がアジアパシフィック (香港)、南米 (サンパウロ)、欧州 (スペイン) の 3 リージョンに拡大しました。事例では Rocket Close が Strands Agents や Bedrock を用いたエージェント型 AI でタイトル業務を最適化した取り組みが紹介されました。セキュリティ速報では AWS Common Runtime aws-c-http の CVE-2026-12043 (ヒープ二重解放) が公開されました。
生成 AI エージェント: Bedrock AgentCore の Web Search 一般提供、SageMaker AI での Nemotron サーバーレスファインチューニング
リージョン拡大: EC2 I7i のパリ提供、Lightsail の 3 リージョン追加
セキュリティ速報: CVE-2026-12043 (aws-c-http のヒープ二重解放)
As AI agents become more capable, they need access to information beyond a model's training data - to answer questions, retrieve latest facts, and take action grounded in current developments. Today, we're making that easy with the general availability of Web Search on AgentCore. Web Search is a fully managed tool that enables agents to ground responses in current, accurate web knowledge while keeping data residency within your secured AWS environment with zero data egress.
Previously, adding web search to agents on Amazon Bedrock AgentCore required integrating with external search providers, building custom orchestration, managing authentication and billing, and coordinating security and compliance across multiple services. Web Search removes this undifferentiated heavy lifting, enabling developers to focus on building agents.
Web Search is built on Amazon’s proven search infrastructure, informed by years of experience powering agentic search experiences across Alexa+, Amazon Q Business, and Kiro. It uses a multi-source grounding approach, by combining a web index operated by amazon with structured knowledge graph data. Beyond standard web results, this gives agents access to entity data and verified facts, helping them retrieve more relevant and accurate responses than traditional web search alone.
Web Search is optimized for agentic retrieval, returning short high-value excerpts that deliver strong intelligence per token. The tool is exposed as a built-in connector target on AgentCore gateway using the Model Context Protocol (MCP). Your agent sends a natural-language query, and Web Search returns ranked results with relevant snippets, source URLs, titles, and publication dates that the model can reason over to produce a grounded response.
Web Search on AgentCore is generally available today in the AWS Region: US East (N. Virginia). For more information, see the AgentCore documentation or read the AWS News Blog.
AWS is announcing the availability of high performance Storage optimized Amazon EC2 I7i instances in AWS Europe (Paris) region. Powered by 5th Gen Intel Xeon Processors with an all-core turbo frequency of 3.2 GHz, these new instances deliver up to 23% better compute performance and more than 10% better price performance over previous generation I4i instances. Powered by 3rd generation AWS Nitro SSDs, I7i instances offer up to 45TB of NVMe storage with up to 50% better real-time storage performance, up to 50% lower storage I/O latency, and up to 60% lower storage I/O latency variability compared to I4i instances.
I7i instances offer compute and storage performance for x86-based storage optimized instances in Amazon EC2 ideal for I/O intensive and latency-sensitive workloads that demand very high random IOPS performance with real-time latency to access the small to medium size datasets. Additionally, torn write prevention feature support up to 16KB block sizes, enabling customers to eliminate database performance bottlenecks.
I7i instances are available in eleven sizes - nine virtual sizes up to 48xlarge and two bare metal sizes - delivering up to 100Gbps of network bandwidth and 60Gbps of Amazon Elastic Block Store (EBS) bandwidth. To learn more, visit the I7i instances page.
Amazon SageMaker AI now supports serverless model customization for Nvidia Nemotron 3 Nano model using supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT). This is a popular open-weight model from Nvidia with 30B total parameters. In addition to deploying this model on SageMaker AI, you can now adapt it to your specific domains and workflows.
Model customization enables you to tailor these foundation models with your proprietary data, whether that's improving accuracy on domain-specific tasks, aligning outputs with your organization's tone, or enhancing performance on new tasks using your labeled data. With serverless customization, SageMaker AI handles all infrastructure provisioning and training orchestration, so you can focus on your data and evaluation rather than cluster management, and only pay for what you use.
Serverless model customization for Nvidia Nemotron 3 Nano on SageMaker AI is available in US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo), and Europe (Ireland). To get started, navigate to the Models page in Amazon SageMaker Studio to launch a customization job, or use the SageMaker Python SDK for programmatic access. To learn more, see the Amazon SageMaker AI model customization documentation.
Amazon SageMaker AI now supports serverless model customization for NVIDIA Nemotron 3 Nano model using supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT). This is a popular open-weight model from NVIDIA with 30B total parameters. In addition to deploying this model on SageMaker AI, you can now adapt it to your specific domains and workflows.
Model customization enables you to tailor foundation models with your proprietary data, whether that's improving accuracy on domain-specific tasks, aligning outputs with your organization's tone, or enhancing performance on new tasks using your labeled data. With serverless customization, SageMaker AI handles all infrastructure provisioning and training orchestration, so you can focus on your data and evaluation rather than cluster management, and only pay for what you use.
Serverless model customization for NVIDIA Nemotron 3 Nano on SageMaker AI is available in US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo), and Europe (Ireland). To get started, navigate to the Models page in Amazon SageMaker Studio to launch a customization job, or use the SageMaker Python SDK for programmatic access. To learn more, see the Amazon SageMaker AI model customization documentation.
Starting today, Amazon Lightsail is available in three additional AWS Regions: Asia Pacific (Hong Kong), South America (São Paulo), and Europe (Spain). This expansion brings the power and simplicity of Lightsail to customers across new geographies in Asia, South America, and Europe.
With this launch, customers in these geographical regions can now enjoy lower latency and better performance for their applications while meeting local data residency requirements. The new Regions provide access to Lightsail's full range of features including instances that meet your compute needs, from general purpose to compute-optimized and memory-optimized bundles, as well as managed databases, container services, load balancers, and more, all with the same simple, predictable pricing that Lightsail customers love. Startups, small businesses, and developers in these regions can now run their applications closer to their end users with low latency.
Lightsail is available in these AWS Regions: US East (Ohio, N. Virginia), US West (Oregon), Canada (Central), Europe (Frankfurt, Ireland, London, Paris, Spain, Stockholm), Asia Pacific (Hong Kong, Jakarta, Malaysia, Mumbai, Seoul, Singapore, Sydney, Tokyo), South America (São Paulo). To learn more about Regions and Availability Zones for Lightsail, please refer to the documentation.
You can create Lightsail resources in these AWS Regions through the Lightsail Console, AWS Command Line Interface (CLI), and AWS SDKs.
Bulletin ID: 2026-043-AWS
Scope: AWS
Content Type: Important (requires attention)
Publication Date: 06/12/2026 11:45 AM PDT
Description:
AWS Common Runtime aws-c-http is a HTTP client library used by AWS SDKs for handling http requests to AWS services. We identified CVE-2026-12043, an issue where improper handling of HPACK dynamic table size updates in the AWS Common Runtime aws-c-http library might allow a remote actor operating a server to cause memory corruption on a connecting client application, potentially leading to arbitrary code execution, via a crafted sequence of HTTP/2 HEADERS frames.
Impacted versions: aws-c-http >= 0.4.22 AND <= 0.10.15
Exposed in following sdk versions:
- aws-sdk-cpp >= 1.11.41, <= 1.11.814
- aws-sdk-java-v2 >= 2.44.27, <= 2.44.14
Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.
In this post, we explore how Rocket Close built a solution using Strands Agents, large language models (LLMs), Amazon Bedrock, Amazon Bedrock Knowledge Bases, and Model Context Protocol (MCP) tools. We cover solution features, the rationale for the technology stack, lessons learned, and the business impact at Rocket Close.