Nov 28, Paris: AI Health Hack #2 with Alan, Qovery and AWS. Applications close Nov 21

What FedRAMP Authorization Actually Requires From a Software Vendor's Cloud Infrastructure in 2026

FedRAMP in 2026 runs on the Consolidated Rules (CR26), with 20x and Rev5 paths and Classes A to D instead of Low, Moderate and High. Here is what your cloud infrastructure must show an assessor before you can sell SaaS to US federal agencies: authorized IaaS, a defined boundary, FIPS 140 crypto, logging, scanning cadence, baselines and supply chain.

Morgan Perry
Co-founder
NOV 20, 2026 · 21 MIN
What FedRAMP Authorization Actually Requires From a Software Vendor's Cloud Infrastructure in 2026

Key points:

  • FedRAMP authorization attaches to one cloud service offering and its assessment boundary, never to a company. Since the Consolidated Rules for 2026 (CR26), FedRAMP calls the result a "FedRAMP Certification", and the old Low, Moderate and High labels map to Classes B, C and D, with a new Class A for providers new to federal work.
  • There are two paths in 2026. FedRAMP 20x lets you apply directly to FedRAMP and prove security with automated Key Security Indicators (46 of them at Class C). The Rev5 path keeps NIST SP 800-53 baselines and, for the agency route, a signed agency ATO. FedRAMP stops accepting new Rev5 certifications on June 11, 2027.
  • Your infrastructure needs eight things an assessor can verify: a FedRAMP-authorized IaaS underneath, a documented boundary with every information flow, FIPS 140 validated cryptography, personnel and access controls (US persons where your data or contract requires it), centralized tamper-resistant logging with tested incident response, a defined vulnerability scanning cadence, automated configuration baselines, and supply chain monitoring.
  • Every third-party tool that can touch federal data or affect its security is a boundary decision. That includes your CI/CD, observability, ticketing and deployment tooling, and CR26's Minimum Assessment Scope rules make you justify each one.
  • Qovery is not FedRAMP authorized and does not make you compliant. It deploys your workloads into a cloud account or Kubernetes cluster you own, with RBAC and audit logs that help with evidence, but whether the Qovery control plane sits inside or outside your boundary is a decision you make with your assessor.

Qovery · Infrastructure control plane
Agent actions, policy-checked before they run
Learn more

The first time a federal prospect asks "are you FedRAMP authorized?", most SaaS teams answer with their cloud provider's badge. AWS, Azure and Google hold FedRAMP authorizations, so the product must be covered too. It isn't. The provider's authorization covers the provider's share of the controls. Everything you build on top of it, from IAM to logging to the pipeline that ships code, is assessed as yours.

The rules changed a lot in 2026, and most guides online still describe the 2023 process. FedRAMP published the Consolidated Rules for 2026 in June, opened Class A applications on August 3, 2026, and makes the new rules mandatory for everyone on January 1, 2027. This article covers the infrastructure side only: what your cloud setup must look like, and what evidence it must produce, before an agency can use your product.

What does FedRAMP authorization require from a software vendor in 2026?

FedRAMP requires a software vendor to define one cloud service offering, draw an assessment boundary around every resource that handles federal customer data or could affect its security, implement the security requirements of a certification class on that boundary, have it assessed, and then keep proving it with continuous monitoring. The cloud infrastructure underneath has to be FedRAMP authorized too.

The legal base is the FedRAMP Authorization Act, enacted on December 23, 2022 as part of the FY2023 NDAA, which gives the General Services Administration (GSA) responsibility for running the program. OMB then issued Memorandum M-24-15 on July 25, 2024, which, in FedRAMP's own words, "formally rescinded and replaced FedRAMP in its entirety, effectively creating a new program with the same name."

CR26 is the operational result. A few definitions from the FedRAMP definitions page matter for anyone scoping infrastructure:

  • Cloud service offering: "a specific, packaged cloud computing product or service supplied by a cloud service provider for use by customers, that is the subject of a FedRAMP Certification." You certify a product, not a company.
  • Minimum assessment scope: "the boundary of the assessment and FedRAMP Certification for the cloud service offering, which effectively defines the cloud service offering itself." This is what older documents called the authorization boundary.
  • Federal customer data: "all electronic information, content, and materials that an agency or its authorized users upload, store, or otherwise supply to a cloud service."
  • Assessor: FedRAMP describes itself as "the final assessor for FedRAMP Certification", with FedRAMP-recognized independent assessors (formerly 3PAOs) doing most of the hands-on work.

Here is how the old vocabulary maps to the new one, based on the FedRAMP Marketplace designations page and notice NTC-0009:

CR26 classMarketplace labelHistorical equivalentTypical fit
Class AStarterFedRAMP ReadyProviders new to federal work; program path only
Class BFoundationRev5 LowLow-impact data, light use
Class CAdvancedRev5 ModerateMost enterprise SaaS used across agencies
Class DHigh AssuranceRev5 HighHigh-impact systems; described as limited to agency-sponsored paths

FedRAMP is explicit that "these classes indicate the depth of information provided in your certification package, not the 'security level' of your product." In practice, the class still follows the impact of the data you handle, because agencies categorize their information under FIPS 199 and pick services certified at a matching level.

Should you go through FedRAMP 20x or Rev5, and do you need an agency sponsor?

For most SaaS vendors starting in late 2026, FedRAMP 20x is the direction of travel. You apply to FedRAMP directly (the "program" path), no agency sponsor required, and you prove security through automated, machine-readable evidence. Rev5 remains open for now, and is still the route for High through an agency, but FedRAMP stops accepting new Rev5 certifications on June 11, 2027.

The FedRAMP 20x page puts it bluntly: "The pilots are over. FedRAMP 20x is here to stay." Phase 2 issued the first Moderate pilot authorizations in March 2026, Phase 3 is active, and Class D (High) under 20x is planned for Phase 4, estimated for FY27 Q1 to Q2.

FedRAMP 20xFedRAMP Rev5
Classes available in late 2026A, B, C (D planned for Phase 4)B, C, D (agency path), plus temporary B and C pipelines
Agency sponsorNot required, you apply to FedRAMPRequired for the agency path, with a signed ATO letter sent to FedRAMP (rule FRC-APS-ATO)
What you proveKey Security Indicators, validated with automationNIST SP 800-53 Rev. 5 baseline controls, documented in an SSP
Evidence formatMachine-readable across the whole scopeSemi-structured text for A to C, full machine-readable data for D from November 1, 2027
Marketplace labelFedRAMP Validated (20x)FedRAMP Certified (Rev5)
Deadline to knowRules mandatory January 1, 2027No new Rev5 certifications after June 11, 2027

"FedRAMP Ready" became legacy on July 28, 2026, so that badge is no longer available as a sales asset, and Class B and C applications opened on August 31, 2026 (CR26 timeline).

If you have no agency sponsor today and your data is Moderate or below, I'd plan for 20x. If an agency is already sponsoring you at High, Rev5 is still the route, and you should expect the evidence bar to rise with the machine-readable requirements.

Which controls apply at each impact level?

At Rev5, the controls come from the FedRAMP baselines built on NIST SP 800-53 Rev. 5: 156 controls at Low, 323 at Moderate and 410 at High, according to the Cloud Security Alliance's breakdown of the baselines FedRAMP released on May 30, 2023. Under 20x, the Class C Key Security Indicators condense those outcomes into 46 indicators across 10 families.

SP 800-53 Rev. 5 has 20 control families, from Access Control to Supply Chain Risk Management. NIST issued release 5.2.0 on August 27, 2025, adding controls on software update integrity (SA-24 and SI-02(07)), so check which release your assessor works from.

The 46 Class C KSIs break down like this:

KSI familyIndicators at Class CInfrastructure examples
Cloud Native Architecture8Restrict inbound and outbound traffic, logical network segmentation
Service Configuration8Automated config management with drift review, key and secret rotation
Identity and Access Management6Passwordless or phishing-resistant MFA, automated account management
Monitoring, Logging, and Auditing5Centralized tamper-resistant logging, persistent log review
Policy and Inventory5Up-to-date inventory of information resources
Change Management4Changes through redeployment of version-controlled resources
Recovery Planning4Backups and tested recovery
Incident Response3After-action reports, reviewed procedures
Supply Chain Risk2Monitor third-party software for upstream vulnerabilities
Cybersecurity Education1Training

Six of the ten families describe how you build and operate infrastructure, and several indicators assume a modern setup. KSI-CMT-RMV expects changes to be "executed through redeployment of version controlled resources rather than direct modification." If your production servers are patched by hand over SSH, that indicator is a problem.

Does the cloud infrastructure underneath your SaaS need to be FedRAMP authorized?

Yes. If your product runs on someone else's infrastructure, that IaaS or PaaS must itself hold a FedRAMP authorization at the same level or higher, so you can inherit its physical, environmental and infrastructure controls. You then document the split in a Customer Responsibility Matrix and implement everything the provider marks as yours.

Authorizations attach to named services in named regions, so verify each service you use on the FedRAMP Marketplace, not on the provider's marketing page. The three hyperscalers each offer a government-focused environment:

EnvironmentWhat the provider statesPersonnel and data locationWhat changes for your team
AWS GovCloud (US)Two isolated US regions (US-East and US-West), marketed for FedRAMP High, DoD and ITAR workloads"Operated by U.S. citizens on U.S. soil"; account holders must be screened US personsSeparate partition with its own IAM system and credentials, separate AWS Organization
Azure GovernmentPhysically isolated US datacenters and networks; three regions (US Gov Arizona, Texas, Virginia)Contractual US data storage, access limited to screened US personsEligibility validation, separate cloud and tenant, some services differ from global Azure
Google Cloud Assured WorkloadsControl packages for FedRAMP Moderate, FedRAMP High, IL2, IL4, IL5 and ITARFedRAMP High package requires first- and second-level support personnel located in the US with enhanced background checksYou configure the folder and package; ITAR requires customer-managed encryption keys

You do not always need a sovereign region. Some commercial-region services hold FedRAMP Moderate authorizations, and for a Class C SaaS with no export-controlled data that can be enough. The sovereign environments earn their extra cost and friction when your data or contract adds requirements FedRAMP alone does not, such as ITAR, DoD impact levels or US-persons-only support.

If you also sell to the defense industrial base, our breakdown of what CMMC 2.0 Level 2 requires from software vendors on a hyperscaler cloud covers the DFARS rule that ties CMMC to FedRAMP Moderate equivalency.

What goes inside the FedRAMP boundary, and what can stay outside?

Everything that handles federal customer data, or is "likely to impact the confidentiality, integrity, or availability" of it, goes inside the boundary under rule MAS-CSO-IIR. That includes the infrastructure that runs your app, the tools that change it, and the tools that see its data.

CR26 adds three rules that change how you scope a modern stack:

  • MAS-CSO-FLO: you must "clearly identify, document, and explain information flows and security categories for ALL information resources." Every arrow in your architecture diagram needs an explanation.
  • MAS-CSO-TPR: for each third-party information resource, you document its usage, why you need it, and the mitigations and compensating controls that reduce impact on federal data.
  • MAS-CSO-MDI: metadata, including metadata about federal customer data, is in scope whenever the resource itself is.

SaaS teams usually get the production VPC right. But the CI runner that builds the image, the deployment tool that holds the cluster credentials, the APM service that receives stack traces, the ticketing system where support pastes customer records, and the secrets manager all have a path to federal data or to the systems that hold it. Each one is either inside your boundary, a FedRAMP-authorized service you inherit from, or a third-party resource you justify under MAS-CSO-TPR.

The cheapest boundary is the smallest one. Teams that pass faster usually run a separate federal environment (its own accounts or subscription, its own cluster) with the same deployment workflow as their commercial product, and they keep commercial SaaS tooling out of it unless it is FedRAMP authorized or carries no federal data.

What will the assessor check in your cloud infrastructure?

The assessor checks that each infrastructure control works and produces evidence, not that a policy document describes it. Under 20x that evidence should come from automation. Under Rev5 it comes from the SSP, scans, configurations and samples. This table maps the infrastructure checklist to both paths.

RequirementRev5 anchor (SP 800-53)20x anchor (Class C KSI)Evidence an assessor expects
Authorized IaaS underneathCA-3, SA-9, inherited controlsMinimum Assessment Scope rulesMarketplace listing per service, Customer Responsibility Matrix
Documented boundary and flowsCA-3, PL-2, SA-9MAS-CSO-IIR, MAS-CSO-FLOBoundary diagram, data flow diagram, inventory
FIPS 140 validated cryptographySC-8(1), SC-13, SC-28(1)KSI-SVC-ASM, KSI-SVC-VRICMVP certificate numbers for each module, TLS and storage configs
Access control and personnelAC-2, IA-2, PS-3KSI-IAM-APM, KSI-IAM-AAMMFA configuration, account lifecycle automation, screening records
Logging and monitoringAU-2, AU-6, AU-9, SI-4KSI-MLA-OSM, KSI-MLA-LET, KSI-MLA-RVLSIEM, list of logged event types, review records
Incident responseIR-4, IR-6, IR-8KSI-INR-RIR, KSI-INR-AARTested runbook, after-action reports, reporting contacts
Vulnerability scanningRA-5, SI-2VDR rulesScan cadence and results, remediation tracking
Configuration baselinesCM-2, CM-6, CM-7KSI-SVC-ACM, KSI-CMT-RMVHardened images, drift detection, infrastructure as code
Network restrictionSC-7KSI-CNA-RNT, KSI-CNA-ULNSecurity groups, network policies, egress rules
Supply chainSR family, SA-24 (5.2.0)KSI-SCR-MON, KSI-SCR-MITDependency and image scanning, provenance, third-party list

FIPS 140 validated cryptography

Every cryptographic function that protects federal data must use a module validated by NIST's Cryptographic Module Validation Program. FedRAMP's Policy for Cryptographic Module Selection and Use, approved by the FedRAMP Board on January 16, 2025, sets this requirement and lets providers track unvalidated modules through POA&Ms while they move to validated ones.

The date to watch passed two months ago. NIST's FIPS 140-3 transition page says FIPS 140-2 modules could be used for new systems until September 21, 2026, after which their certificates move to the Historical list and remain acceptable only for existing systems. If you are building a federal environment now, look for FIPS 140-3 certificates.

This touches the TLS terminator on your load balancer, the ingress controller in your cluster, the base images of your containers (the OpenSSL build matters), disk and object storage encryption, and the KMS that holds your keys. Cloud provider APIs often offer FIPS endpoints, but the containers you build are your problem.

US persons and personnel controls

The FedRAMP baselines require personnel screening and access control, but US-citizens-only operations usually come from the data or the contract, not from FedRAMP itself. ITAR data, DoD impact levels and some agency contracts require that only US persons administer the system or access its data.

That requirement reaches your vendors too. If a support engineer at one of your providers can open a shell on a production node, that person's location and screening is part of your answer. This is why sovereign environments advertise US-persons support, and why every tool with production access needs a line in your boundary documentation.

Logging and incident response

You need centralized, tamper-resistant logging and evidence that someone reviews it. KSI-MLA-OSM asks for a SIEM "or similar system(s)" for "centralized, tamper-resistant logging", KSI-MLA-LET for a maintained list of what you log, and KSI-MLA-RVL for logs that are "persistently reviewed and audited."

Incident reporting got stricter with CR26. Reporting deadlines are now tiered by class and incident severity rating (N1 to N5). According to Davis Wright Tremaine's summary of the Consolidated Rules, the shortest deadline is 15 minutes for an initial report of an N3, N4 or N5 event in a Class D system, and the longest is one business day for lower-severity incidents in lower classes. Fifteen minutes means detection, triage and a named on-call person have to work at 3 a.m., which is an infrastructure problem as much as a process one.

Vulnerability scanning cadence

CR26's Vulnerability Detection and Response rules set scanning and remediation by class. At Class C, rule VDR-TFR-PDD says providers should run vulnerability detection on resources likely to drift at least once every 14 days. That drops to every 7 days at Class D.

Remediation timeframes depend on the vulnerability's potential agency impact rating (PAIN), its likely exploitability and its internet reachability. At Class C, the most severe category, a likely exploitable and internet-reachable PAIN-5 vulnerability, should be mitigated within 2 days of evaluation. Existing certified services have to be fully migrated to these rules by December 7, 2026, with a grace period to March 7, 2027, according to FedRAMP's VDR migration page.

Two days rules out a monthly patch window. You need image scanning on every build, fast rebuilds and redeploys, and a record of when each fix reached production.

Configuration baselines

Configuration must be automated and checked for drift. KSI-SVC-ACM asks that configuration be "managed using automation and persistently reviewed for drift," and KSI-CMT-RMV wants changes made by redeploying version-controlled resources. On the Rev5 side, CM-2 and CM-6 expect documented baselines (typically CIS Benchmarks or DISA STIGs) for every component type.

Containers and infrastructure as code make this easier to prove. A hardened base image, a Kubernetes cluster defined in Terraform, and deployments that only happen through a pipeline give you a baseline, drift detection and a change log in one place. Hand-edited production does the opposite.

Supply chain

FedRAMP expects you to know what third-party software you run and to watch it. KSI-SCR-MON requires that "third party software information resources are automatically monitored for upstream vulnerabilities," and KSI-SCR-MIT that supply chain risks are "persistently identified, reviewed, and mitigated." NIST's release 5.2.0 added SA-24 on software update integrity.

For a SaaS team that means dependency scanning, container image scanning, pinned and signed base images, builds that run in an environment you control, and a list of every third-party service with a justification, which is the same list MAS-CSO-TPR asks for.

Planning a federal environment on infrastructure you own?
Qovery deploys your apps on Kubernetes in your own cloud account, with RBAC and audit logs your assessor can read. Qovery is not FedRAMP authorized, so let's walk through your boundary together before you commit.

Who assesses you, and what happens after you are authorized?

A FedRAMP-recognized independent assessor tests your implementation, FedRAMP makes the final certification decision, and on the agency path the sponsoring agency also signs an ATO. After that, you keep the certification alive through continuous monitoring, annual assessment and change reporting.

What continuous monitoring means in 2026:

  • Persistent validation (20x) replaces the yearly manual snapshot with automated checks of your KSIs.
  • Annual independent assessment still applies; rule IVV-CSX-AIA mandates it for 20x Class A, and the other classes have their own assessment rules.
  • Vulnerability detection and response on the cadences above, with reporting to your agency customers.
  • Significant Change Notifications, mandatory for all Rev5 services from January 1, 2027. Moving to a new region, swapping your container platform or adding a third-party tool inside the boundary can all count.
  • Incident communications on the tiered deadlines.

On time and cost, I won't quote a number. FedRAMP does not publish official figures, and the estimates floating around vendor blogs vary a lot and mostly predate CR26. The levers you control are the size of your boundary, how much evidence your infrastructure produces on its own, and whether you start with a sponsor or on the program path.

Where does Qovery fit in a FedRAMP environment, and where doesn't it?

Qovery is not FedRAMP authorized, holds no FedRAMP certification at any class, and using it does not make your product compliant. Qovery holds SOC 2 Type II, which is a different framework. What Qovery can do is run your deployments on infrastructure you own and produce access and change records that help with evidence.

Qovery is an Agentic DevOps platform with a split architecture. The Qovery control plane (Console, API, orchestration) is a commercial SaaS. Your applications run on Kubernetes in your own cloud account (AWS, GCP, Azure or Scaleway) or on a Kubernetes cluster you already run, connected with the Qovery Operator or qovery cluster install. Qovery never hosts your application or its data.

Here is what that gives a team building a federal environment:

  • Workloads stay in your account. The federal cluster, its VPC, its KMS keys and its storage sit in an account you own, so you inherit your IaaS provider's authorization directly. With the Qovery Operator on an existing cluster, images are built inside your cluster and pushed to a registry in your own cloud account (BYOK docs), which keeps the build in your environment.
  • One workflow for commercial and federal. The same git-push deployments, environment cloning and Terraform provider work in both, so the federal environment does not become a second, hand-maintained platform.
  • Evidence for access and change. Role-based access control across the organization and its projects, and audit logs that record who deployed or changed what, map to AC-2, AU-2 and CM-3 style evidence. API Policy Tokens, built on OPA, scope exactly what an automation or AI agent may do, and each agent's actions show up in the audit log under its own token name (how OPA policies work in Qovery).

Three points decide whether Qovery belongs in your boundary at all, and you settle them with your independent assessor:

  • Information flows to the control plane. The Operator opens an outbound connection from your cluster to Qovery. Deployment metadata, service configuration, application logs streamed to the Console depending on your setup, and any environment variables or secrets you store in Qovery (rather than in a secrets manager integration) cross that connection. Under MAS-CSO-FLO and MAS-CSO-TPR, each of those flows needs documentation, and your assessor decides whether any of it counts as federal customer data or affects its security.
  • Privileged access. Qovery deployment Jobs run inside your cluster with cluster-admin rights. That makes the Qovery control plane a resource "likely to impact" the integrity of your system, which is exactly what the Minimum Assessment Scope rules ask you to account for.
  • Partition and personnel specifics. Sovereign environments such as AWS GovCloud (US) use a separate partition with its own IAM system. Whether a given Qovery setup fits your specific government partition, and whether a non-FedRAMP control plane is acceptable to your agency at all, are questions to settle in a proof of concept before you write anything into an SSP.
OptionWhere workloads runControl plane FedRAMP statusBoundary impactBest fit
Hand-built pipeline inside the boundary (self-hosted CI plus kubectl or Argo CD)Your accountInside your boundary, assessed as yoursNo external party, but you build and prove everythingTeams with a platform team and time
A FedRAMP-authorized DevOps or CI/CD serviceYour account or the vendor'sAuthorized, check the Marketplace listingInherit its authorization and CRMTeams that want an authorized tool and accept its constraints
Qovery on a cloud account or cluster you ownYour accountNot FedRAMP authorizedThird-party resource or out-of-boundary, decided with your assessorTeams that want one workflow and own-account workloads, and can justify the control plane or keep it out of scope
Run a commercial-only product and decline federal dealsYour accountNot applicableNoneTeams for whom the federal market does not justify the cost yet

I'd pick the hand-built option if an agency tells you no external control plane is acceptable, and I'd pick an authorized tool if your assessor wants every dependency on the Marketplace. Qovery fits best where the federal environment is one of several you run, the boundary design keeps federal data away from the control plane, and your assessor agrees with the justification. If you are thinking about isolation beyond FedRAMP, our piece on running AI air-gapped on your own Kubernetes covers the far end of that spectrum. More frameworks are mapped on our compliance topic page, including NIST CSF-aligned deployment platforms.

How do I set this up with Qovery from Claude Code or Cursor?

Start outside the federal boundary. Use a commercial or staging environment that mirrors your federal architecture, let your coding agent deploy and operate it through Qovery, and bring the resulting design (accounts, cluster, flows, access model) to your assessor before anything touches federal data.

BASH|1. Install the Qovery Agent Skills for Claude Code, Cursor, Codex, OpenCode, Gemini CLI
curl -fsSL https://skill.qovery.com/install.sh | bash

# 2. Or install them for the current project only
curl -fsSL https://skill.qovery.com/install.sh | bash -s -- --project

# 3. Add the Qovery MCP Server to Claude Code (OAuth, read-only by default)
claude mcp add --transport http qovery https://mcp.qovery.com/mcp --callback-port 4242

# 4. Write access, only when you explicitly want the agent to act (also enable it in the Qovery Console)
claude mcp add --transport http qovery "https://mcp.qovery.com/mcp?read_write=true" --callback-port 4242

Step 1 or 2 installs the skills; you approve the install script. Step 3 opens an OAuth sign-in you approve in the browser, and the agent can only read. Step 4 is the only one that lets the agent change things, and it needs write access enabled in the Console too. Then ask in plain language, for example "deploy this app to my staging AWS account with Qovery" or "list who changed the staging environment this week." The setup guides are the AI agent quickstart and the Claude Code integration.

Frequently asked questions
What does FedRAMP authorization actually require from a software vendor's cloud infrastructure before they can sell to US federal agencies?

It requires a FedRAMP-authorized IaaS or PaaS underneath, a documented assessment boundary covering every resource and information flow that touches federal customer data, and the controls of your certification class implemented and evidenced on that boundary. In infrastructure terms that means FIPS 140 validated cryptography, strong identity and personnel controls, centralized tamper-resistant logging, tested incident response, scanning on a set cadence (every 14 days for drifting resources at Class C), automated configuration baselines and supply chain monitoring. An independent assessor tests it, FedRAMP certifies it, and continuous monitoring keeps it valid.

What is the difference between FedRAMP 20x and FedRAMP Rev5 in 2026?

FedRAMP 20x lets you apply directly to FedRAMP without an agency sponsor and proves security with automated Key Security Indicators (46 at Class C). Rev5 uses the NIST SP 800-53 Rev. 5 baselines, with an agency ATO on the agency path. Both now follow the Consolidated Rules for 2026, and FedRAMP stops accepting new Rev5 certifications on June 11, 2027.

Do I need AWS GovCloud or Azure Government to get FedRAMP authorized?

Not always. You need IaaS services authorized at your class or higher, and some commercial-region services hold FedRAMP Moderate authorizations, so check each service on the FedRAMP Marketplace. AWS GovCloud (US), Azure Government and Google Cloud Assured Workloads become necessary when your data or contract adds requirements such as ITAR, DoD impact levels or US-persons-only support.

Does FedRAMP require FIPS 140-3 instead of FIPS 140-2?

FedRAMP requires FIPS 140 validated modules for cryptography that protects federal data. NIST moved FIPS 140-2 certificates to the Historical list after September 21, 2026, so they remain acceptable only for existing systems, and a new federal environment should use FIPS 140-3 validated modules.

Is Qovery FedRAMP authorized, and does using it make my product compliant?

No to both. Qovery holds SOC 2 Type II but no FedRAMP authorization, and FedRAMP certification attaches to your cloud service offering and its boundary. Qovery deploys your workloads into a cloud account or Kubernetes cluster you own and gives you RBAC and audit logs for evidence, but you and your independent assessor decide whether the Qovery control plane sits inside your boundary as a justified third-party resource or stays outside it.

Can my CI/CD and observability tools stay outside the FedRAMP boundary?

Only if they cannot reach federal customer data and cannot affect the security of the systems that hold it. CR26's Minimum Assessment Scope rules require you to document every information flow and justify every third-party resource. A SaaS APM that receives request payloads, or a CI system holding production credentials, is almost always in scope unless it is FedRAMP authorized or you redesign the flow.

Morgan Perry
About the author
Morgan Perry

Morgan co-founded Qovery and leads engineering. He writes about Kubernetes architecture, DevOps best practices, and building resilient infrastructure at scale.

Next step

Planning a federal environment on infrastructure you own?

Qovery deploys your apps on Kubernetes in your own cloud account, with RBAC and audit logs your assessor can read. Qovery is not FedRAMP authorized, so let's walk through your boundary together before you commit.