AWS Weekly Roundup: Claude Fable 5.1 Lands on Bedrock, Amazon Linux 2027 Preview
AWS's September 7, 2026 weekly roundup centered on Claude Fable 5.1 becoming available through both Amazon Bedrock and the Claude Platform on AWS. Anthropic positions the model for "ambitious tasks across coding, scientific research, and enterprise workflows" and extended work sessions that span multiple applications. As a Covered Model, Fable 5.1 on Bedrock comes with data retention policies (up to 30 days) and human safety reviews through a new aws_review data retention mode — and eligible customers can instead keep data inside their own cloud environment via Anthropic's Enterprise Frontier Safeguards, which began rolling out this month.
The roundup also covered the public preview of Amazon Linux 2027 (AL2027), the next iteration of AWS's Linux distribution. Built on kernel 7.1+ and extending AL2023's foundation, AL2027 targets cloud-native workloads — web applications, databases, containerized microservices, AI/ML workloads, and large-scale infrastructure — with an emphasis on performance and security hardening at the OS layer.
AWS also opened beta registration for a new AWS Certified AI Business Strategist certification, aimed at non-technical roles — line-of-business managers, sales professionals, consultants, and program managers — who need to evaluate and champion AI adoption without necessarily writing code themselves. Beta exam registration opened September 1, 2026, with formal exam delivery starting September 29, 2026.
Read more — AWS News Blog
Amazon EBS Volume Clones Now Work Across AWS Accounts
AWS extended EBS Volume Clones — previously limited to copying volumes within the same Availability Zone and account — to support cross-account copying. The workflow is a two-step handoff: the volume owner shares the EBS volume through AWS Resource Access Manager (RAM), and the target account accepts the resource share, then selects "Copy volume" for any shared volume that appears in their EBS console.
For volumes encrypted with a customer-managed KMS key (CMK), the target account can specify a different CMK to re-encrypt the copy under its own keys during the clone — useful for teams that want cryptographic isolation between production and downstream accounts. Volumes encrypted with AWS-managed keys can't be shared this way, since AWS-managed keys aren't exportable to other accounts by design.
The practical use case AWS highlights is refreshing test and development environments with current production data without exposing the production account itself: a secondary account can pull a clone of the latest volume state to develop and experiment against, while the production data stays isolated in its own account boundary.
Read more — AWS News Blog
Google Cloud Brings Real-Time AI Inference to Pub/Sub
Google Cloud's Pub/Sub team shipped AI Inference Single Message Transforms (SMTs), which enrich Pub/Sub messages with model predictions from Gemini Enterprise Agent Platform in real time, without standing up a separate enrichment service. Developers configure the SMT on either a topic or a subscription, pointing it at an Agent Platform endpoint (self-deployed or Model-as-a-Service), and the SMT merges the model's JSON response into the outgoing message alongside the original payload. Google recommends attaching SMTs at the subscription level rather than the topic level, since subscription-level SMTs can lean on Pub/Sub's existing retry and dead-letter-topic handling if the model endpoint is temporarily unavailable. Typical use cases include real-time sentiment analysis, content classification, embedding generation, and cutting latency out of AI-enrichment pipelines that previously required a dedicated consumer service.
In a related rebrand, Google Cloud renamed the legacy BigQuery insertAll streaming API to the BigQuery Storage Write API (REST), preserving 100% backward compatibility for existing callers. The REST surface remains the pick for serverless apps, IoT telemetry, and AI logging use cases that want simple HTTP-based streaming, while the gRPC version of the Storage Write API stays the recommended choice for high-throughput, continuous ingestion pipelines.
Read more — Google Cloud Documentation
AlloyDB Omni's RPM Orchestrator Reaches GA for Bare-Metal PostgreSQL
Google Cloud announced general availability of the AlloyDB Omni Red Hat RPM Orchestrator on September 10, 2026, bringing production-grade automation to AlloyDB Omni deployments running on bare-metal servers and VMs rather than in Kubernetes or as a standalone container. The orchestrator supports four deployment modes — standalone container, container with a Kubernetes operator, standalone RPM, and RPM orchestrator for high availability — and the HA mode adds automatic failover, synchronous replication across zones, and read pools for scaling out read traffic.
Operationally, the orchestrator automates low-downtime maintenance, dynamic configuration tuning, and backup/restore to either GCS or S3 with point-in-time recovery, plus SELinux enforcement and local audit logging for teams with strict compliance requirements. AlloyDB Omni claims roughly 2x the transactional throughput and up to 100x faster analytical queries compared to standard PostgreSQL, and the RPM orchestrator extends that performance profile — along with vector search, natural-language queries, and direct Gemini model access — to on-premises and edge environments that can't run in Google Cloud proper. For regulated industries with data-residency constraints, this closes a real gap: the same AI-ready Postgres-compatible database, automated the same way, whether it's running in GCP or in a data center you control.
Read more — Google Cloud Blog
Links & Sources
You May Also Enjoy
