Is ChatGPT FedRAMP Authorized? What Agencies Need to Know

The certified ChatGPT offering includes ChatGPT Enterprise and the OpenAI API Platform. The other ChatGPT products are excluded, including the free tier, Plus, Team, and the separately announced ChatGPT Gov. The Federal Risk and Authorization Management Program (FedRAMP) Marketplace product record lists the certified offering as "ChatGPT Enterprise and API Platform," Package identification (ID) FR2533155773, at FedRAMP 20x Class C (Moderate).

Agency staff who read "ChatGPT is FedRAMP authorized" and provision the wrong product can place Controlled Unclassified Information (CUI) into a service absent from the FedRAMP AI product list. That misstep runs against the scope rules in Office of Management and Budget (OMB) Memorandum M-24-15 (July 25, 2024) and, at many agencies, against the agency generative AI policy.

Key Takeaways

• Enterprise and API only. The listing is FedRAMP 20x Class C (Moderate), and use requires the gov.api.openai.com endpoint.
• ChatGPT Gov is self-hosted. Agencies host it in their own Azure tenant and issue an agency-specific authorization to operate for it.
• Impact level sets boundaries. The Moderate certification covers CUI; High-categorized data needs ChatGPT Gov or the Amazon Web Services (AWS) GovCloud Bedrock path.
• Agency authorization remains required. Each agency issues its own Authority to Operate (ATO) and reviews continuous monitoring deliverables.

ChatGPT Enterprise and the OpenAI API Platform Hold FedRAMP 20x Certification

ChatGPT Enterprise and the OpenAI API Platform hold FedRAMP certification through the FedRAMP 20x path. The Marketplace records a Certified Since date of January 9, 2026, and the Moderate-level certification was announced publicly on April 27, 2026.

Access inside the boundary runs through a dedicated government endpoint. FedRAMP compliance requires that "all API requests use the designated gov.api.openai.com endpoint for all supported methods and models."

The listing's vocabulary comes from the Consolidated Rules for 2026, released June 24, 2026, with mandatory adoption for all stakeholders beginning January 1, 2027. The current ruleset is v2026.07.14.01.

Under the terminology changes, "FedRAMP Authorized" became "FedRAMP Certified," and providers now elect a Certification Class from A through D. FedRAMP prints the legacy impact-level label in parentheses after the class through December 31, 2026, which is why the entry reads "20x Moderate (Class C)" rather than a bare impact level.

That managed-service boundary doesn't extend to agency-hosted ChatGPT Gov deployments.

ChatGPT Gov Is a Separate Deployment Agencies Host Themselves

ChatGPT Gov was announced on January 28, 2025, as a tailored version of ChatGPT that gives US agencies another way to reach OpenAI's frontier models. Agencies run it exclusively inside their own Microsoft Azure commercial or Azure Government tenant on top of Azure OpenAI Service. The agency keeps operational control of the environment; with ChatGPT Enterprise, OpenAI keeps it.

That hosting choice is what lets ChatGPT Gov reach environments the Moderate certification doesn't cover. Azure Government holds a FedRAMP High Provisional Authority to Operate (P-ATO) and Impact Level 2, 4, and 5 (IL-2, IL-4, IL-5) provisional authorizations. That deployment model lets agencies meet IL-5, Criminal Justice Information Services (CJIS), International Traffic in Arms Regulations (ITAR), and FedRAMP High requirements.

ChatGPT Gov has no standalone FedRAMP Marketplace listing, so an agency pursues its own internal authorization for the deployment and uses Azure Government's existing authorizations.

That control costs IT investment. The agency operates the tenant and runs the administrative console for single sign-on and user management. It also owns the security, privacy, and compliance work. OpenAI performs that work for Enterprise customers.

The deployment distinction also leaves consumer and Team products outside any authorized boundary.

Consumer ChatGPT Tiers Fall Outside the Authorized Boundary

The Marketplace entry names one product, "ChatGPT Enterprise and API Platform." Consumer ChatGPT at ChatGPT.com, in its free and Plus tiers, and ChatGPT Team appear nowhere. The prioritization criteria required five enterprise capabilities:

• Enterprise identity.
• Provisioning.
• Access control.
• Analytics features.
• Guaranteed data separation.

The consumer tiers provide none of them. A shared product name doesn't extend a certification across product lines.

The prohibition follows from FedRAMP's scope. The Consolidated Rules for 2026 require agencies to promote the use of FedRAMP-certified cloud products and services when their use is within scope, and OMB M-24-15 defines that scope as cloud services that "create, collect, process, store, or maintain Federal information on behalf of a Federal agency."

FedRAMP's AI scope examples draw the line: an agency using a shared cloud service to index and search internal repositories of non-public federal information is in scope, while developers using an AI coding assistant "on entirely public code repositories" are out of scope. A free ChatGPT account that receives CUI, personnel records, legal documents, or internal drafts falls on the in-scope side and has no certification.

The scope logic is not new; several agencies acted on it well before OpenAI held any certification.

Agency Restrictions on Public AI Preceded OpenAI's Certification

Before OpenAI held any certification, the General Services Administration (GSA) blocked generative AI on government devices in June 2023. Other agencies followed:

• Department of Veterans Affairs (VA). Restricted ChatGPT access in July 2023.
• Department of Homeland Security (DHS). Banned public AI tools available to the general public in May 2025.

Each restriction preceded OpenAI's certification, and none has been lifted purely on the basis of the FedRAMP listing. Each agency still applies its own policy on top of the certification.

For certified or agency-authorized products, the next constraint is the data's impact level.

Impact Level Determines What Data an Agency Can Process with ChatGPT

FIPS 199 categorizes federal systems as Low, Moderate, or High by the worst-case effect of losing confidentiality, integrity, or availability. It sets the system's impact level at the highest of the three security objectives, and FIPS 200 then uses that categorization to define minimum security requirements.

Federal guidance distinguishes the categories as follows:

• Low. Covers publicly available or non-sensitive information.
• Moderate. Covers serious operational or financial damage short of loss of life, including CUI such as personnel records and legal documents tied to internal agency operations.
• High. Covers systems whose compromise threatens life or financial ruin, typically law enforcement, emergency services, financial, and health systems.

Federal impact-level examples show that, in practice, a platform managing agency budget data is Moderate; a law enforcement case-management system or an emergency dispatch platform is High.

Under the Consolidated Rules for 2026, FedRAMP no longer ties a provider's control set to an impact level. Providers elect a Certification Class, and Class C is meant for common enterprise services used across an agency.

Agencies still categorize their own systems under FIPS 199 and select controls from the NIST SP 800-53 Rev5 catalog, and the Class C (Moderate) ChatGPT Enterprise service is assessed for controlled unclassified information workloads.

An agency with High-categorized data has two OpenAI paths today: ChatGPT Gov on Azure Government, or the AWS GovCloud Bedrock OpenAI API, which is listed at FedRAMP High and IL-5. FedRAMP's high-impact Class D pilot is planned for Phase 4 (fiscal year 2027, quarters 1 to 2), so no 20x High path exists yet for the managed product.

Other AI Tools Already Hold FedRAMP Certification on the Marketplace

The AI Prioritization Initiative is closed to new entrants, so its early 2026 roster is complete even as the wider Marketplace grows. Three services define the current field:

• ChatGPT Enterprise and API Platform (OpenAI). Package FR2533155773, FedRAMP 20x Class C (Moderate), with the Moderate-level announcement dated April 27, 2026.
• Gemini for Government (Google). FedRAMP 20x Moderate, with FedRAMP High-authorized security and compliance features. Available under a GSA OneGov agreement.
• Perplexity Enterprise and API Platform (Perplexity AI, Inc.). Class B (Low); available to agencies through the GSA Multiple Award Schedule.

The current roster is short, and each entry carries distinct class, impact-level, and procurement details. Confirming status is only the first step; each agency still runs a five-part workflow before any authorized AI tool reaches production.

Agencies Must Complete Five Steps Before Deploying Any FedRAMP AI Tool

Nicole Thompson, security director of FedRAMP at the GSA, stated the rule in May 2026: agency authorization "is always, and has always been, required to use a FedRAMP-authorized product at an agency." A provider's certification is a government-wide security assessment; the agency ATO is agency risk acceptance.

Agencies must follow five steps.

1. Confirmation of FedRAMP Certified Status

Under FedRAMP status guidance, FedRAMP Ready can't support an ATO decision, and In Process means the provider is still working toward certification and isn't yet certified. For ChatGPT Enterprise, agencies review the shared-responsibility expectations and validation evidence in OpenAI's Trust Portal.

2. FIPS 199 Categorization Match

The agency makes the scope determination and decides whether its data is Low, Moderate, or High. A Class C service assessed for Moderate-categorized use supports CUI workloads; High-categorized data needs a different path.

3. Agency ATO Issuance Through Package Reuse

OMB M-24-15 requires that "agencies must presume the security assessment documented in the authorization package is adequate for their use in issuing an authorization to operate at or below that FIPS 199 impact level."

Under agency ATO requirements, the ATO must identify the approved use, information boundary, required configurations, restrictions, and conditions for continued use, and the agency notifies FedRAMP after issuing it.

4. FedRAMP Workspace and Endpoint Provisioning

Existing commercial ChatGPT Enterprise workspaces can't be converted. OpenAI supports a one-time migration of users and conversations into a newly provisioned FedRAMP workspace, with all API traffic routed to gov.api.openai.com.

5. Continuous Monitoring for the ATO Lifecycle

Continuous monitoring runs for as long as the ATO remains in effect. The M-24-15 presumption of adequacy holds only while the provider maintains its ongoing requirements.Under the Consolidated Rules for 2026, providers use the vulnerability model, while agencies maintain remediation plans. Each consuming agency also approves deviation and significant change requests and reviews the annual Third-Party Assessment Organization (3PAO) assessment; those post-ATO requirements can't be delegated.

The good news is agencies can compress this sequence. A service deployed inside an already-authorized infrastructure boundary can inherit a significant share of the required NIST SP 800-53 Rev5 controls, which lets agency review begin against a partially validated environment rather than a greenfield package. That pre-authorized boundary pattern shifts where certification time is spent, but it does not remove the agency's ATO or continuous monitoring obligations.

SaaS Vendors Ready for Certification Can Move to Market Now

Agency authorization work now sets the pace of federal AI adoption. With the AI Prioritization Initiative closed and OpenAI, Google, and Perplexity already certified, the bottleneck for new agency AI deployments has shifted from provider assessment to agency ATO, monitoring cadence, and boundary-inheritance choices. Providers that reach the Marketplace early sit in front of buyers whose remaining work is on the agency side.

Knox Systems operates a pre-authorized FedRAMP boundary across AWS, Azure, and Google Cloud Platform (GCP), giving SaaS vendors a predictable path to authorization in roughly 90 days at about 90% less cost than the traditional route. The platform currently supports FedRAMP Moderate, FedRAMP High, and Defense Information Systems Agency (DISA) IL-4, with IL-5 authorization in process and estimated for December 2026.

Vendors preparing for federal adoption can book a demo to scope a certification path.

FAQs about ChatGPT FedRAMP Authorization

Is ChatGPT FedRAMP Approved?

Yes. The FedRAMP configuration adds access restrictions and feature controls while retaining Enterprise policies, including the Compliance API.

Can Federal Employees Use ChatGPT on Government Systems?

Availability depends on agency implementation and policy, not an employee's federal status alone. Under the GSA-OpenAI partnership, GSA made ChatGPT Enterprise available to the executive branch for $1 per agency in August 2025, but each agency determines whether and how employees may use it.

Which AI Tools Are FedRAMP Approved for Government Agencies?

Anthropic's Claude for Government is FedRAMP High authorized. Infrastructure-level authorizations can cover AI products that do not appear on the AI prioritization page.

Can the Government Access or View ChatGPT Conversations?

OpenAI limits employee access to conversations to incident resolution and cases where applicable law requires it. Employees may also recover end-user conversations with explicit permission. Workspace administrators can pull a conversation audit log through the Enterprise Compliance API, and qualifying organizations can set a zero data retention policy on the API platform.

‍