FedRAMP and IL levels for AI SaaS in defense

A FedRAMP label and a DoD Impact Level are not the same scale. FedRAMP said so when it dropped “Impact Levels” from its own vocabulary. Official DoD pages still use IL2, IL4, IL5 and IL6 for cloud information. This page keeps those systems apart and records what the cited pages say about AI SaaS.

Status as of 14 October 2026: the Consolidated Rules for 2026 still take effect on July 4, 2026. The same page now also says only a few Rev5 rulesets are mandatory prior to January 1, 2027 and that enforcement of the Consolidated Rules for 2026 will begin after that date. FedRAMP’s AI page still dates the 2025 AI Prioritization Initiative August 2025 through April 2026 and still says This opportunity is no longer available for new entrants. ChatGPT Enterprise, Gemini for Government and Perplexity Enterprise Pro for Government still Received FedRAMP Certification in early 2026. The War CIO AI Cybersecurity Risk Management Tailoring Guide (14 July 2025) still states the IL2 / IL4 reciprocity in the same FedRAMP-authorization wording. DoD CIO CMMC About and the FedRAMP equivalency PDF 403ed this pass.

FedRAMP Certification is not a government-wide ATO

The Consolidated Rules for 2026 take effect on 4 July 2026. The same page says only a few Rev5 rulesets are mandatory prior to January 1, 2027, and that After January 1, 2027, enforcement of the Consolidated Rules for 2026 will begin. FedRAMP replaced the historical term “FedRAMP authorization” with “FedRAMP Certification” so that a program certification is not confused with an agency authorization to operate.

The same table replaced “Impact Levels / Low, Moderate, High” with “Certification Class / A, B, C, D”. Agencies are required by FIPS 200 to separately and uniquely identify the target Impact Level and control baselines for each federal information system.

RFC-0020’s outcome notice is explicit about DoD vocabulary: FedRAMP will not use the term “levels” or numbers for the new baseline labels to avoid confusion with the DOD/DOW Impact Level/IL system. A FedRAMP Certification is not a guarantee that a cloud service has met all requirements to be appropriate for use by an agency at a given FIPS 199 security category. FedRAMP does not have the authority to make this determination on behalf of an agency authorizing official.

Those sentences already answer the common overstatement “FedRAMP Moderate equals DoD IL4.” FedRAMP’s own pages refuse that mapping.

What FedRAMP published about AI cloud services

FedRAMP’s AI page says the 2025 AI Prioritization Initiative began in August 2025 and was completed in April 2026. This opportunity is no longer available for new entrants.

The historical criteria were for AI-based cloud services that provide access to conversational AI engines designed for routine and repeated use by federal workers, including enterprise features, data separation and protection so that any model information from training on customer data will not leave the customer environment without customer authorization, demand from at least five CFO Act agencies, GSA Schedule availability, and readiness for a FedRAMP 20x authorization.

The same page lists three services that Received FedRAMP Certification in early 2026: ChatGPT Enterprise and API Platform (OpenAI); Gemini for Government (Google); Perplexity Enterprise Pro for Government. That is a historical prioritization list, not a DoD Impact Level catalog, and not a claim that those services are authorized for CUI or classified information.

FedRAMP will stop accepting any new FedRAMP Rev5 Certifications on June 11, 2027. Existing Rev5 certifications will remain active until at least December 31, 2028.

What official DoD pages say about IL and FedRAMP

The Department of War CIO AI Cybersecurity Risk Management Tailoring Guide (14 July 2025) is the official page we found that states the IL mapping in AI terms:

• A CSO with a Federal Risk and Authorization Management Program (FedRAMP) authorization equates to a DoD Impact Level 2 (IL2) (i.e., public information) authorization through reciprocity.
• A CSO adopting a FedRAMP authorization for DoD missions at a non-public level (i.e., Controlled Unclassified Information (CUI) or non-CUI) must also adopt FedRAMP+ controls specifically tailored to DoD. This would equate to a DoD Impact Level 4 (IL4).

The same guide tells readers to see the DoD Cloud Computing Cloud Service Provider SRG and DoD Cloud Computing Mission Owner SRG for higher impact levels and FedRAMP+ controls. It also says a DISA issued provisional authorization (PA) can be leveraged by a mission-owner authorizing official for an ATO.

DoD CIO’s FedRAMP Authorization and Equivalency pamphlet is the CMMC-program page this directory already cites for cloud CUI. It is about CMMC cloud equivalency, not a substitute for an IL4 or IL5 provisional authorization. See CMMC and AI with CUI.

About CMMC is the live status page for Phase I / Phase II. It does not, on the text we cite, rename FedRAMP Certification as an Impact Level.

What we searched for and did not find

We searched the FedRAMP AI page and the 2026 rules page for a sentence that assigns IL4, IL5 or IL6 to a named AI SaaS. We did not find one. We also searched the War CIO tailoring guide for a sentence that says a FedRAMP Certification Class (A–D) equals a DoD IL. We did not find that either. The guide still uses the older word “FedRAMP authorization” when it states the IL2 / IL4 reciprocity.

Our reading, not an official table: public or non-CUI DoD use of a FedRAMP-certified conversational AI service is the IL2 fact pattern those pages describe. CUI in an AI SaaS is the FedRAMP+ / IL4 fact pattern. Classified (IL5 / IL6) is pointed to the Cloud Computing SRGs, not to fedramp.gov/ai. A CMMC cloud equivalency letter is a third instrument.

This is general information, not an authorization. The mission-owner authorizing official still issues the ATO.