Claude FedRAMP authorisation is one of the most common questions we receive from federal agencies, government contractors, and defence organisations evaluating AI adoption. The direct answer as of early 2026: Anthropic is actively pursuing FedRAMP authorisation for Claude, and Claude is available on AWS GovCloud through Amazon Bedrock — which carries FedRAMP High authorisation. For many federal use cases, this is the viable deployment path today.
This guide covers the current authorisation landscape, the architectural options available to government organisations right now, the specific security requirements that apply to government Claude deployments, and the procurement considerations that defence contractors and civilian agency teams face. The regulatory environment for government AI is evolving rapidly — what follows reflects the situation as of March 2026.
Key Takeaways
- Claude is available via AWS Bedrock on GovCloud, which holds FedRAMP High authorisation — the primary path for federal deployments today
- Anthropic is pursuing direct FedRAMP authorisation; check the FedRAMP Marketplace for current status
- NIST SP 800-53 controls apply to federal AI deployments — your system security plan must address how Claude fits your security boundary
- Executive Order 14110 on Safe, Secure, and Trustworthy AI applies to federal agency AI deployments
- SB 1047-equivalent state legislation and EU AI Act may apply to government contractors with international operations
- Our Security & Governance service supports FedRAMP system security plan development
FedRAMP Current Status and the AWS GovCloud Path
FedRAMP (Federal Risk and Authorization Management Program) is the US government's standardised security assessment framework for cloud services used by federal agencies. A FedRAMP authorisation means a cloud service has been assessed against NIST SP 800-53 controls and approved for use by federal agencies — saving each agency from conducting its own assessment. Without FedRAMP authorisation, federal agencies generally cannot use commercial cloud services for federal information.
As of March 2026, the primary path to using Claude with federal information is through Amazon Bedrock on AWS GovCloud. AWS GovCloud (US) holds FedRAMP High authorisation — the highest level, required for systems handling the most sensitive federal non-classified data. Claude models are available through Amazon Bedrock's model catalogue, and because Bedrock runs within the FedRAMP-authorised GovCloud environment, agencies can inherit the infrastructure authorisation when building Claude-powered applications in that environment.
FedRAMP authorisation statuses change. Before making procurement or architecture decisions, verify the current authorisation status for both Bedrock and Claude specifically in the official FedRAMP Marketplace at marketplace.fedramp.gov. This guide reflects the situation as of March 2026.
Using Claude through Bedrock on GovCloud means your application is built on FedRAMP-authorised infrastructure, but you still need to document your own security controls, complete your Authority to Operate (ATO) process, and maintain your system security plan (SSP). The infrastructure authorisation doesn't extend to the applications you build on it.
Deployment Options for Government Organisations
Government organisations evaluating Claude have several deployment paths, with different security postures, data handling properties, and procurement implications. Choosing the right path depends on your data classification requirements, existing cloud infrastructure, and the specific use case.
NIST SP 800-53 Controls for Claude Deployments
Federal AI deployments must address applicable NIST SP 800-53 controls in their system security plans. While the specific controls required depend on your system's impact level (Low, Moderate, High) and the data it processes, several control families are particularly relevant to Claude deployments.
Access Control (AC)
NIST access control requirements for federal systems require that access to Claude-powered applications be managed through your agency's identity management infrastructure. This means integrating Claude applications with PIV/CAC card authentication, implementing least-privilege access controls, separating duties for privileged functions, and implementing session management controls including automatic logout. The access control requirements are more prescriptive for federal systems than for commercial enterprise — plan for this in your architecture from the start.
Audit and Accountability (AU)
Federal systems must generate, protect, and retain audit records sufficient to enable the monitoring, analysis, investigation, and reporting of unlawful or unauthorised system activity. For Claude applications, this means the comprehensive audit logging described in our Claude audit logging guide, but with federal-specific requirements: audit records must be protected from modification and unauthorised access, audit review must be conducted at defined frequencies, and audit record retention must meet federal requirements (typically a minimum of 3 years, with specific requirements depending on your system category).
Configuration Management (CM)
Federal CM controls require maintaining baseline configurations for your Claude applications, documenting changes, analysing security impacts before implementing changes, and maintaining documentation of the software components in your system — including Claude model versions. Model version changes from Anthropic need to be treated as configuration changes and assessed for security impact before being applied to production systems.
Incident Response (IR)
Federal incident response requirements are more prescriptive than commercial standards: agencies must report security incidents to US-CERT (now CISA) within defined timeframes, maintain incident handling capability, and test incident response procedures regularly. Your incident response plan for Claude applications must integrate with your agency's broader IR capability and address AI-specific scenarios including model misbehaviour, prompt injection attacks, and data exfiltration through AI interfaces.
Government AI Architecture Review
We help federal agencies and government contractors design Claude deployments that meet FedRAMP, NIST 800-53, and CMMC requirements — with full documentation for your ATO package.
Book a Government AI Consultation →Executive Order 14110 and Federal AI Policy
Executive Order 14110 on Safe, Secure, and Trustworthy AI, issued in October 2023, establishes requirements for federal agency AI use. Federal agencies deploying AI systems — including Claude — must adhere to OMB's guidance implementing this order, which includes requirements for AI impact assessments, transparency, bias evaluation, and human oversight for consequential AI decisions.
OMB Memorandum M-24-10 (Advancing Governance, Innovation, and Risk Management for Agency Use of Artificial Intelligence) established specific requirements for federal agencies using AI in high-impact use cases. These include designating a Chief AI Officer, developing AI governance structures, conducting pre-deployment risk assessments, maintaining inventories of AI use cases, and ensuring human oversight for consequential decisions. If you're deploying Claude in a federal agency context, your CAIO needs to be aware of the deployment and it needs to be included in your AI use case inventory.
The NIST AI Risk Management Framework (AI RMF) is increasingly referenced in federal AI governance. It defines four core functions for AI risk management: Govern, Map, Measure, and Manage. Federal agencies deploying Claude are expected to implement AI risk management practices aligned to this framework. Our responsible AI framework guide covers the NIST AI RMF in the context of Claude deployments.
Defence Contractors: CMMC and Claude
Defence contractors subject to the Cybersecurity Maturity Model Certification (CMMC) programme face specific requirements for handling Controlled Unclassified Information (CUI) and Federal Contract Information (FCI). CMMC 2.0 has three levels: Level 1 (basic cyber hygiene, self-assessed), Level 2 (advanced practices, third-party assessment for most CUI environments), and Level 3 (expert practices, government-led assessment).
If your Claude deployment processes CUI or FCI, it falls within your CMMC compliance boundary and must implement the appropriate NIST SP 800-171 practices (for Level 2) or additional NIST SP 800-172 practices (for Level 3). Claude processing CUI for a defence application is not a free lunch — it expands your compliance boundary and the assessment scope. Engage your legal and compliance teams and your CMMC Registered Practitioner Organisation (RPO) before building Claude applications that handle CUI.
The practical implication for most defence contractors: deploy Claude through AWS GovCloud or a similarly accredited environment, implement access controls through your CAC-authenticated identity infrastructure, implement comprehensive audit logging, and ensure your subprocessor (AWS or Anthropic) agreements address CUI handling requirements. We have experience designing Claude architectures for defence contractors — contact us for a confidential discussion of your specific requirements.
State and Local Government Deployments
State, local, and education (SLED) organisations are not subject to FedRAMP requirements directly, but many have adopted state-level equivalents or require cloud services to meet FedRAMP standards as a proxy for their own assessments. Before deploying Claude in a SLED context, confirm what your state's cloud security requirements are — they vary significantly.
State and local government use cases for Claude are compelling: policy analysis, constituent service automation, document processing, procurement support, and administrative workflow automation. For government Claude deployments at the state and local level, the primary considerations are often data residency (many states have data sovereignty requirements), integration with legacy systems (MCP servers can bridge Claude to existing government databases), and procurement (many states require open competitive procurement for AI services above specific thresholds).
Claude Enterprise's administrative controls — usage limits, user management, activity monitoring — align well with the oversight requirements many state CIOs implement for AI systems. If your state has an AI policy or AI use guidance, review it before deployment and ensure your Claude deployment complies with any categorisation or approval requirements it establishes.
Government Procurement Considerations
Government procurement for AI services differs meaningfully from commercial enterprise purchasing. FAR and DFARS clauses, competition requirements, data rights provisions, and small business set-aside considerations all apply to AI procurements. Anthropic and its resellers participate in government procurement channels — check GSA Schedule availability and SEWP for existing contract vehicles that may allow you to procure Claude services without a full open competition.
Data rights in government AI contracts require attention. Your government contract may give the government unlimited rights to data created under the contract — understand how that interacts with AI training data provisions in your agreements with Anthropic. Anthropic's enterprise agreements explicitly state that customer data is not used to train models — that provision is important for your government contracting data rights analysis.
For IT modernisation projects that include Claude, ensure your Independent Government Cost Estimate (IGCE) includes ongoing API costs, which scale with usage and differ from the per-seat pricing of traditional software. Token-based pricing can be difficult to estimate without baseline usage data. Start with a bounded pilot to establish usage baselines before committing to larger-scale deployment contracts.