In its weekly summary, published on September 7, 2026, AWS announced the availability of the Claude Fable 5.1 model on its platform, alongside opening the public preview of Amazon Linux 2027 and launching updates to computing, data, and AI agent management services. The article brings together multiple announcements, but its main focus is on running long-duration AI tasks and the controls AWS imposes on certain sensitive models.
Claude Fable 5.1 Arrives on AWS with Additional Controls
According to Anthropic, Claude Fable 5.1 was designed to handle tasks lasting hours, including programming, scientific research, and enterprise workflows. It can execute larger parts of software projects, including working across the codebase, reviewing code, and improving performance during long sessions.
Anthropic classified the model as a Covered Model, a category subject to additional policies for data retention, security review, and access control. AWS says Claude Fable 5.1 may be subject to data retention for up to 30 days and human review by Amazon employees. It also added the aws_review mode, which allows prompts and outputs to be retained for human security review within AWS boundaries. The provider_data_share mode is legacy, and the article explains that Amazon Bedrock does not share user data with the model provider.
The model can be accessed through Amazon Bedrock or Claude Platform on AWS. Enterprise Frontier Safeguards, developed by AWS and Anthropic, also allows eligible customers to use covered models while keeping their data in a cloud environment they control.
Infrastructure and Data Updates
Amazon Linux 2027 has entered public preview. It is the next version of Amazon Linux and runs on kernel 7.1 or later. It is based on AL2023, with a clear focus on cloud-native workloads, including web applications, databases, containerized microservices, and artificial intelligence and machine learning workloads.
AWS also announced the memory-optimized Amazon EC2 R9g and R9gd instances, powered by AWS Graviton5 processors. The company cites up to a 25% improvement in compute performance compared with Graviton4-based R8g and R8gd instances, along with gains of up to 30% for databases, 35% for web applications, and 35% for machine learning.
Lambda SnapStart is also now available for container image functions, after previously being limited to the managed Python, .NET, and Java runtimes. AWS says the feature may reduce startup time from several seconds to less than one second in cases such as machine learning inference and interactive application programming interfaces.
What Changes in Practice?
These announcements present two parallel paths: increasing the ability of AWS services to run long and complex workloads, and tightening governance around models that require review and data retention. However, benefiting from Claude Fable 5.1 is not separate from retention and review policies, which is an important constraint for teams handling sensitive data.
Other updates included the general availability of AWS Agent Registry as a private, governed catalog for agents, tools, MCP servers, and custom resources, with search, approval workflows, and AWS CloudTrail audit logs. Amazon Redshift also added support for Apache Iceberg v3 tables, including column default values, row sequencing, and deletion vectors, on equipped Graviton clusters and the Redshift Serverless service.
New Certification for Business Professionals
AWS opened registration for the beta version of the AWS Certified AI Business Strategist certification on September 1, 2026, with the exam scheduled to begin on September 29. The certification targets business unit leaders, sales professionals, consultants, and program managers who evaluate AI initiatives, support their adoption, and connect investment in them to business outcomes.