New updates on AWS GovCloud (US) | Issue #5 (2026)
July 16–31, 2026 | New updates on AWS GovCloud (US): This issue features insights on FedRAMP Consolidated Rules 2026 (CR26), new foundation models now available on Amazon Bedrock in AWS GovCloud (US) including Claude Sonnet 5, Grok 4.3, and Gemma 4, model migration guidance for retiring models, plus the latest features and services.
Executive Insights
What the New FedRAMP Consolidated Rules for 2026 (CR26) Mean for AWS GovCloud (US) Customers
FedRAMP Consolidated Rules for 2026 (CR26) went into effect on July 4, 2026, for early adoption, ahead of mandatory adoption on January 1, 2027. The rules change how cloud services are approved for AWS GovCloud (US) customers with FedRAMP compliance needs, but do not affect customers with DoD CC SRG IL-4 and IL-5 compliance needs. As part of this change, FedRAMP now designates certifications by class rather than impact level, with Classes A through D indicating increasing depth of security evidence. FedRAMP High approval is now expressed as Class D certification, and AWS GovCloud (US) has adopted the Class D (High) certification. All existing FedRAMP authorizations carry forward under the new class designations. This applies to AWS services authorized on AWS GovCloud (US), so the AWS services customers use today remain FedRAMP certified. The new process replaces documentation-heavy, point-in-time assessments with continuous, automated validation of machine-readable Key Security Indicators (KSIs). AWS has aligned its services to this automation-first model of assurance. For agencies, continuous validation provides current visibility into provider security posture. Machine-readable Significant Change Notifications (SCNs) add visibility into service changes as they happen, with notice requirements scaled to the significance of the change. For technology partners, automation can reduce authorization timelines to a matter of weeks, and those building on AWS GovCloud (US) inherit AWS security controls and can automate KSI evidence collection with services such as AWS Audit Manager, AWS Config, and AWS Security Hub. For industrial base members, faster authorizations and a lower documentation burden mean new capabilities reach federal missions sooner. Organizations planning a federal authorization should factor the new rules into their compliance roadmaps now. To learn more, including how to request the AWS FedRAMP Security Package or a FedRAMP Partner Package, visit AWS FedRAMP Compliance and Prepare for FedRAMP 20x with AWS automation and validation on the AWS Public Sector Blog, or reach out to your AWS account team.
Resources: AWS FedRAMP Compliance | Prepare for FedRAMP 20x with AWS automation and validation | AWS DoD SRG Compliance | FedRAMP CR26 Timeline | FedRAMP Certification Classes | FedRAMP Significant Change Notification | FedRAMP Marketplace: AWS GovCloud
Solutions Insights
Newer, More Capable Models Are Here — and Upgrading Is Straightforward
Here's what practitioners need to know about the latest model launches and how to upgrade from retiring models on Amazon Bedrock in AWS GovCloud (US).
Here's what practitioners need to know about the latest model launches and how to upgrade from retiring models on Amazon Bedrock in AWS GovCloud (US).
Three new foundation models are now available in AWS GovCloud (US), bringing the authorized model catalog to 20+ models spanning Amazon, Anthropic, Meta, NVIDIA, OpenAI, Google DeepMind, and xAI.
-
Claude Sonnet 5 (Anthropic) is now available in both AWS GovCloud (US) Regions via cross-Region inference on the runtime endpoint and on the Bedrock Mantle endpoint in AWS GovCloud (US-West), with FedRAMP Class D and DoD CC SRG IL-4/5 authorization. Sonnet 5 is Anthropic's most capable Sonnet model, built for coding, agents, and professional knowledge work, handling large codebases and multi-file changes with fewer correction rounds while holding state reliably across long tool-using workflows.
-
Grok 4.3 (xAI) is available in AWS GovCloud (US-West) on the Bedrock Mantle endpoint with FedRAMP Class D and DoD CC SRG IL-4/5 authorization, offering Standard, Priority, and Flex service tiers so teams can match cost and throughput to the workload using in-Region inference.
-
Gemma 4 (Google DeepMind) is an open-weight (Apache 2.0) family, including 31B dense, 26B-A4B mixture-of-experts, and E2B variants, with built-in reasoning, native function calling, multimodal (text + image) input, and a 256K-token context window. Gemma 4 is available in AWS GovCloud (US-West) on the Bedrock Mantle endpoint, with FedRAMP Class D and DoD CC SRG IL-4/5 authorization in progress, and is a strong fit for multimodal agents, document-understanding pipelines, and cost-efficient software workflows.
Whether the workload calls for frontier reasoning, cost-efficient inference, multimodal document understanding, or high-throughput embeddings, practitioners can match the right model to the mission.
With the breadth of models now available in AWS GovCloud (US), teams running earlier-generation models have a clear path forward. Upgrades are a deliberate model-ID change you control, not an automatic switch, and newer models improve both capability and price-performance. Bedrock's model lifecycle provides at least 12 months of availability for every model, followed by at least 6 months in Legacy state before end-of-life, with notifications along the way. That said, Claude Sonnet 3.5 and 3.7 retired on July 30, so teams still calling these models should migrate immediately. The next retirement is Claude 3 Haiku on September 10, 2026.
The migration process is straightforward, and the tools to validate it run natively in AWS GovCloud (US). Start by inventorying what you run, checking the modelLifecycle field via ListFoundationModels/GetFoundationModel or the Bedrock console. Then pick your target. Before you switch, validate by running your candidate side-by-side with your current model using Bedrock Evaluations, available in both us-gov-west-1 and us-gov-east-1 alongside batch inference, so you can benchmark automatically, with human review, or using LLM-as-a-judge before you cut over. Take the opportunity to re-tune prompts for the new model's strengths. Once validated, ship by updating the model ID in your applications, agents, and Knowledge Bases. Enabling a new model starts in your linked commercial account (us-east-1 or us-west-2) by establishing the model agreement. Then enable the same model on the Model Access page in your AWS GovCloud (US) Region. Entitlements propagate in a few minutes.
Learn more: Bedrock model support by Region | Understanding the Amazon Bedrock model lifecycle | Bedrock Evaluations|LLM-as-a-judge best practices | Optimize and migrate prompts in Amazon Bedrock | Bedrock in AWS GovCloud (US)
Service and Feature Releases
Source: What's New Posts, July 16–31, 2026.
Helpful Links
Did you find what you were looking for today?
Let us know so we can improve the quality of the content on our pages