September 7, 2026, (Inside AI) — Amazon Web Services has made Anthropic’s Claude Fable 5.1 available on its platform, a model the AI lab describes as built for long-running, high-stakes work that spans hours and many applications. The release marks another step in the deepening operational relationship between AWS and Anthropic, with new data governance features that go beyond typical model hosting arrangements.
Claude Fable 5.1 is designated by Anthropic as a Covered Model, a category that carries additional data retention, safety review, and access policies wherever the model is offered. On AWS, the model is subject to data retention for up to 30 days and human review by Amazon personnel, with a new data retention mode. In this mode, AWS retains prompts and outputs for human safety review within the AWS boundary. The mode is legacy, and Amazon Bedrock does not share data with the model provider.
Enterprise Frontier Safeguards (EFS), built in partnership between AWS and Anthropic, will let eligible customers use Covered Models while keeping their data in a cloud environment they control. The safeguards address a persistent enterprise concern: how to use frontier models without exposing proprietary data to third parties. AWS has not disclosed which customers qualify for EFS or the technical implementation details.
Customers can access Claude Fable 5.1 through Amazon Bedrock and Claude Platform on AWS. The model card on Amazon Bedrock provides technical specifications and usage guidance. Anthropic positions Fable 5.1 as capable of owning more of a software project on its own, handling features across an entire codebase, code review, and performance work over extended sessions. That framing targets software engineering teams looking to delegate complex, multi-step tasks to an autonomous system.
Covered Models Introduce New Governance Layer for Enterprise AI
The Covered Model designation is a relatively new concept from Anthropic, introduced as part of its enterprise safety framework. It signals that certain models require stricter oversight because of their advanced capabilities or potential for misuse. For AWS customers, the practical effect is a shift in data handling expectations. The 30-day retention window and human review by Amazon personnel may conflict with strict data residency or privacy policies in regulated industries such as healthcare and finance.
Inside AI could not independently verify whether the human review process is automated or manual, or how Amazon selects reviewers. AWS has not published a detailed data flow diagram for the retention mode. The company did state that Amazon Bedrock does not share data with Anthropic, a meaningful distinction from other cloud providers that may pass data to model developers for safety review. The EFS program appears designed to give enterprise customers more control, but its eligibility criteria remain opaque.
Industry analysts note that data governance is becoming a competitive differentiator in the AI cloud market. Microsoft Azure and Google Cloud have both introduced similar enterprise controls for frontier models, though their approaches vary. AWS’s partnership with Anthropic gives it a unique position, as Anthropic has historically emphasized safety and alignment more than some competitors. The Covered Model framework may appeal to risk-averse enterprises.
Read: Amazon Bedrock Adds OpenAI GPT-5.6 Models with India-Only Inference
AWS Expands Infrastructure and Certification Portfolio
Beyond the Claude Fable 5.1 rollout, AWS announced several other updates. Amazon Linux 2027 (AL2027) entered public preview, running on kernel 7.1+ and purpose-built for cloud-native workloads. The operating system builds on AL2023’s baseline and targets web applications, databases, containerized microservices, AI/ML workloads, and large-scale infrastructure. AWS said AL2027 is designed for customers who need a secure, stable, and AWS-native operating system.
Amazon EC2 R9g and R9gd memory-optimized instances launched, powered by AWS Graviton5 processors. AWS claims up to 25% better compute performance compared to Graviton4-based R8g and R8gd instances, with up to 30% faster database performance and up to 35% faster web application and machine learning performance. These numbers are vendor benchmarks and have not been independently validated.
AWS Lambda SnapStart now supports container image functions, reducing startup times from several seconds to as low as sub-second for latency-sensitive workloads such as ML inference and interactive APIs. Previously, SnapStart was only available for managed runtimes including Python, .NET, and Java. The AWS Agent Registry became generally available, providing a private, governed catalog for agents, tools, skills, MCP servers, and custom resources. New enterprise features include manual and URL-based record creation, approval workflows, semantic and keyword search, and AWS CloudTrail audit trails.
Amazon Redshift added support for Apache Iceberg v3 tables, including default column values, row lineage, and deletion vectors. The feature works on Graviton-based provisioned and serverless clusters. AWS also announced a new certification, AWS Certified AI Business Strategist, targeting professionals who evaluate and scale AI initiatives. Beta exam registration opened September 1, 2026, with exam delivery beginning September 29.
For the 16th consecutive year, Gartner recognized AWS as a Leader in the 2026 Magic Quadrant for Strategic Cloud Platform Services, placing AWS highest on the Ability to Execute axis. AWS also collaborated with the SANS Institute on a chapter in the 2026 Cloud Security Exchange eBook, outlining a framework for securing agentic workloads at enterprise scale. The chapter covers architectural patterns and implementation guidance for security teams at every stage of agentic AI maturity.
