Anthropic announced that its Claude models have been authorized for use in FedRAMP High and Department of Defense (DoD) Impact Level 4 and 5 (IL4/5) workloads via Amazon Bedrock, available in AWS GovCloud (US) regions. The announcement was made on June 11, 2025.
What the authorization means
FedRAMP High and DoD IL4/5 are among the highest cloud security certifications for handling unclassified and controlled unclassified information. With this approval, federal agencies and defense organizations can deploy Claude models in sensitive but unclassified mission contexts that require stringent government security controls.
Available models and Bedrock features
At launch the following Claude variants are available:
- Claude 3.5 Sonnet v1
- Claude 3 Haiku
These models are offered through Amazon Bedrock alongside platform capabilities such as:
- Amazon Bedrock Agents (to automate multistep tasks)
- Guardrails (customizable safety and compliance controls)
- Knowledge Bases (integration of external knowledge stores)
- Model Evaluation (tools to assess and select appropriate models)
Anthropic stated additional Claude models may be added later, including Claude 3.7 Sonnet and Claude 4.
Use cases for government customers
Anthropic highlighted use cases for federal and defense customers, including:
- Deploying frontier AI for complex document analysis and intelligence synthesis in secure environments
- Building AI agents that process controlled unclassified information
- Leveraging Claude’s advanced reasoning and a 200,000-token context window to analyze large datasets comprehensively
Accelerating government AI adoption and multi-platform flexibility
Anthropic’s stated goal is commercial availability of Claude models across the regions where public sector customers operate, while meeting required security accreditations for mission-critical workloads. Claude on Amazon Bedrock is provided as a fully managed service intended to reduce infrastructure complexity while preserving necessary security controls.
The announcement also notes government customers now have flexibility to deploy capable AI systems across multiple secure platforms: for example, Google Cloud Vertex AI for FedRAMP High and IL2, and Amazon Bedrock for FedRAMP Moderate and High and DoD IL2 and IL4/5.
How to get started
Federal agency employees, defense contractors, and approved partners can access Claude models through Amazon Bedrock in AWS GovCloud (US) regions using familiar Amazon Web Services APIs and management tools. Anthropic’s public sector team can be contacted for guidance on implementing Claude in federal environments and ensuring compliance with government security requirements.
Additional notes
The announcement does not include pricing details or a specific timeline for adding future models beyond the statement that additional variants may be introduced later.
