AWS News - 2026-09-14

AWS End User Messaging now supports dynamic WhatsApp flows so you can turn WhatsApp conversations into rich, interactive experiences that end-users complete directly in WhatsApp.

Instead of pushing end-users to a website or a separate form to complete an interactive experience, you can instead guide them through an appointment booking, lead generation, sign-up, survey or a transaction directly in the chat. Customers can start building a flow using one of Meta's stater templates with familiar building blocks like text inputs, date pickers, drop downs, and buttons, or customize the flow from scratch. With dynamic flows, each screen can call your own HTTPS endpoint in real time, so you can show live availability, and personalize each step in the flow for your end-users.

You build and manage flows entirely in AWS End User Messaging Social, through the console or APIs. Define a flow with a JSON schema, then deliver it with a message template to reach customers where they are.

WhatsApp flows are available in all regions where End User Messaging Social is available. To get started, see the User Guide.

AWS End User Messaging is advancing its deliverability capabilities in phone pools. Now, customers sending through phone pools automatically benefit from automatic failover to the next best performing number in the pool when SMS delivery is impacted from downstream disruptions.

SMS deliverability can be affected by drops or delays when trying to reach the end recipient while it's being routed from AWS to messaging providers and eventually mobile carriers. These disruptions can hit the most important parts of your messaging program, including delivery notifications, one-time passcodes, reminders, and other time-sensitive messages. AWS End User Messaging watches for delivery delays, message failures, or drops in conversion and reroutes traffic to a more optimized delivery path.

Customers do not need to take any action to receive this benefit for phone pools. If you are not sending through phone pools today, you can create one by following these instructions. This capability is available in all AWS Regions where AWS End User Messaging is available.

To learn more, see the AWS End User Messaging SMS User Guide.

AWS Glue zero-ETL integrations now detect table property conflicts and track integration ownership. When you configure a source table and target catalog, Glue associates the resulting table properties with the owning integration, so two integrations can no longer be pointed at the same target table without your knowledge. This works across Amazon S3 Tables and SageMaker Lakehouse catalogs.

Data teams running multiple zero-ETL integrations gain predictable control over where each source table lands. If you attempt to create or modify an integration whose table properties are owned by another integration, Glue identifies the owning integration and guides you to choose a different target or update the existing one, so your pipelines stay isolated and your data lands exactly where you intend.

This feature is available in all AWS Commercial and AWS GovCloud (US) Regions where AWS Glue zero-ETL integrations are supported.

To learn more, refer to the AWS Glue Developer Guide and get started in the AWS Glue console.

Amazon S3 Object Lock がイベントホールドによる可変リテンションに対応、Amazon EMR が旧リリースからの移行中のお客様向けに無償でサポート延長、Amazon API Gateway がバックエンド統合向け mutual TLS をサポート、OpenAI GPT-6 Astra が Amazon Bedrock で GA、Amazon Quick デスクトップアプリが macOS と Windows で GA、AWS Lambda Managed Instances が Graviton5 搭載インスタンスと 90 分タイムアウトに対応、Amazon EVS が大阪を含む 4 リージョンに拡大、AWS Elemental Inference がライブ動画からコンテキストメタデータをリアルタイム生成、Amazon API Gateway が 1 MB の実行ログと配信先の設定に対応、AWS DevOps Agent が Slack との双方向通信に対応、AWS Lambda が Amazon S3 Files のダイレクトリード設定に対応等

週刊生成AI with AWS, ようやく秋の気配を感じる2026年9月7日週号 – Umios様 (旧マルハニチロ) の時系列基盤モデルChronos-2による販売計画AI事例、CODASの産業特化型AI基盤の取り組み含む14件のブログ記事を紹介。サービスアップデートでは、OpenAI GPT-6 AstraのAmazon Bedrock一般提供開始、Bedrock Managed Knowledge BaseのTwelveLabs Marengo 3.0対応、Amazon Quickデスクトップアプリの一般提供開始含む18件を紹介。

セキュリティ業務で AI を活用する際、脆弱性を発見できるかだけでなく、安全なコードを誤検出しないかが信頼の鍵になります。本記事では、実際の脆弱性と誤検出を見分ける能力を測る初のベンチマーク Deception Benchmark を紹介します。16 のプログラミング言語、70 を超える CWE カテゴリの 14,822 サンプルで 12 モデルを評価した結果、FPR (偽陽性率) と FNR (偽陰性率) の両方を 10% 未満に抑えられた構成はありませんでした。設計手法と結果、実務への示唆を詳しく解説します。

Amazon Science の研究領域に新たに加わった自動推論について、予備知識のない方に向けてわかりやすく解説します。簡単な C と Python のコード例を使い、網羅的テストでは 1,300 年以上かかる検証を自動推論がミリ秒で示せる理由、テストが今後も必要な理由、停止性問題ゆえに「Don't know」という答えが避けられない事情、論理・定理証明・形式手法といった関連用語の違いを説明します。IAM Access Analyzer や VPC Reachability Analyzer など AWS での活用例と、学習用の書籍・ツールの一覧も紹介します。

英語版ブログ： “ Executive Insights from the Inaugural AWS Lif […]