Anthropic Claude models in AWS GovCloud let teams run AI coding for regulated workloads

Running Anthropic Claude via Amazon Bedrock in AWS GovCloud lets a small team trial AI-assisted coding inside a regional, compliance-oriented environment, but you must budget time to confirm model certifications and update internal controls.

What actually changed

Amazon reports that Anthropic Claude Opus 5.5 and Claude Sonnet 5.5 are now available in AWS GovCloud (US) through Amazon Bedrock and that Anthropic’s agentic coding tool, Claude Code, can be used with those models. The announcement describes this as an "on-ramp for AI-assisted development" for workloads with regulatory or compliance requirements, and explicitly calls out use cases such as ITAR. The post also states that Claude Sonnet 5 holds FedRAMP Class D certification and DoD Impact Level 4 and 5 (IL4/IL5) authorisation; it says Opus 5.5 and Sonnet 5.5 hold FedRAMP Class D certification on Bedrock and recommends verifying current model certification status.

Who it affects

• Government contractors and teams handling regulated data (for example, ITAR-covered work) who need to keep processing inside AWS GovCloud.
• Small engineering teams that want AI assistance for routine development tasks but must keep their workloads within authorised environments.
• Compliance and security owners who will need to validate whether these model placements meet programme requirements.

What it costs or what it replaces

The announcement does not state pricing. Concretely, this gives organisations an option to run AI-assisted development inside a GovCloud boundary rather than routing requests to public-region model endpoints, which can reduce regulatory exposure if the rest of your control set aligns. You should not assume any change to your existing cloud billing or contract terms until you verify Amazon’s pricing and your AWS account entitlements.

What we don't know

• Whether the model certifications and authorisations cited are unchanged at the time you read this (the blog says to verify current model certification status).
• Pricing for Bedrock usage, Claude Code, or model inference in AWS GovCloud — the announcement does not state pricing.
• Exact operational steps, required IAM roles or networking setup to run Bedrock and Claude Code in a GovCloud account.
• Any contractual or data-residency guarantees beyond the region placement described in the post.

What to do next

1. Read the AWS Machine Learning Blog post and share it with your compliance lead. Confirm whether your workload is practised under regulations named in the post (for example ITAR) and whether FedRAMP/DoD IL requirements apply to your programme.

1. Verify current certification status and availability in your target GovCloud region. Ask your AWS account team or check Amazon Bedrock documentation to confirm that the specific model (Opus 5.5 or Sonnet 5.5) and Claude Code are authorised for your use; do not proceed until you have written confirmation.

1. Run a short pilot in AWS GovCloud this week: choose a non-sensitive, routine development task, provision Bedrock access through your existing GovCloud account, and test Claude Code with one repository. Document data flows, logging, and any controls you must add before scaling to sensitive workloads.

1. Based on the pilot, update your compliance checklist and operational runbooks (access controls, audit logging, data handling). Escalate any gaps to your security/compliance owner before moving regulated data or expanding usage.