- FedRAMP told more than 500 certified cloud services to move to new vulnerability rules or lose certification, Endor Labs reports. Mandatory adoption starts December 7, 2026.
- The rules reward documented judgment. Every finding needs a verdict with evidence, and FedRAMP warns that careless 'not exploitable' labels can come back to hurt providers.
- Ask who owns the evaluation step, how fast your team can mitigate, and whether reporting can run without manual assembly.
The scan is no longer the product
Cloud providers that sell to US federal agencies used to be judged on activity. Scan often, log the findings, watch the deadline. The new FedRAMP rules ask for something harder. Each finding needs a reasoned verdict, and the reasoning must be on file.
Endor Labs, a security vendor, laid out the changes on November 20. It says FedRAMP wrote on November 10 to more than 500 certified cloud services. The notice sets a cut-off of March 8, 2027. A provider that has not moved to the new rules by then, and has no approved corrective plan, loses its certification. It is also barred for a year, as Endor Labs quotes the notice.
The lesson here is that freedom to judge comes with a duty to prove. Providers gain room to decide what is truly exploitable. In return, each decision becomes a claim someone can hold them to after an incident.
What changed and why
FedRAMP is the General Services Administration program that certifies cloud services for federal customers. Endor Labs says its monitoring rested on monthly scans, severity scores and fixed repair deadlines. Those deadlines applied whether or not the vulnerable code could ever execute.
Endor Labs links the change to a CISA directive, BOD 26-04, issued in June. It shifts federal patching away from uniform deadlines and toward risk. The factors are exposure, known exploitation, automatability and impact.
FedRAMP then brought forward the mandatory start of two rule sets. They are Vulnerability Detection and Response, and Vulnerability Evaluation and Reporting. The date moved from June 2027 to December 7, 2026.
Endor Labs also cites a J.P. Morgan report that finding vulnerabilities is outpacing fixing them by 16.5 times.
How the new rules work
Providers must judge every finding on several points, Endor Labs explains. Is it likely to be exploited? Could a payload sent from the internet trigger it, even deep inside the software?
Exploits are presumed automatable unless the provider has evidence otherwise. In plain terms, automatable means an attacker could script it and run it at scale. Each finding also gets an impact rating for agencies, from N0 to N5.
Speed matters too. Providers get a short window to assess each finding. Endor Labs puts it at two to fourteen days from detection, set by the provider's certification class. Known exploited flaws follow the BOD 26-04 timelines. Endor Labs notes that federal agencies face the same December 7 date under that directive.
Mitigation now counts. Under a rule called VDR-CSO-RES, a provider can shrink exposure until the flaw is no longer automatable or internet-reachable. A full upgrade can wait. The catch, Endor Labs says, is producing safe mitigations quickly and proving they work.
What a defensible verdict looks like
Take a hypothetical. A scanner flags a software library your product uses. The flaw sits in one function that parses a rare file type. Your application never calls that function.
Writing "not exploitable" is not enough. FedRAMP warns that labelling findings this way "without careful consideration" will cost the provider if one becomes the root cause of an incident.
Better evidence has three parts. First, analysis showing no code path from your application reaches that function. Endor Labs says function-level analysis of this kind is accepted by assessors, including Fortreum and Schellman. Second, a check that no internet input can reach it. Third, a dated record of who decided. Endor Labs says its fixes also produce a VEX record, a short statement of why a flaw cannot be exploited.
Base images are harder. A base image is the pre-built operating layer a container starts from. Endor Labs says most container flaws come from there, not from your own code. Your team did not write it and cannot patch it directly.
Judging those findings means separating them from application findings. It also means knowing whether a newer image would fix or add problems. That is difficult to do inside a window of days if the habit is to rebuild once a quarter.
Where programs will strain
Reporting turns into a data feed. Providers must send a record for each vulnerability in FedRAMP's JSON format. They must also send monthly readable reports, and every finding carries a 192-day clock. Endor Labs argues that no team can stitch this together by hand each time a deadline arrives.
Matt Bruggeman of A-LIGN told Federal News Network that many providers' current tooling falls short of the new requirements. As a result, he said, some are searching for new providers.
Treat the vendor's numbers with care. Endor Labs sells the tools it recommends. It says reachability analysis cuts open-source findings by 92% on average. It says only 9% of Egnyte's findings proved reachable. In its research on 264,643 reachable findings, only 10% could be fixed by a simple upgrade. For 82%, a fix existed in first-party code. These are the company's own figures, and the article does not describe its method.
Questions to put to your team
Start with a name. Who decides that a finding is not exploitable, and who signs that decision? Endor Labs argues that clear ownership of this step matters as much as any tool.
Next, ask for a dry run. Can the team produce per-finding records and a monthly report without a scramble? How many days does it take to mitigate a flaw in a base image?
Then ask what to fund. The sources give no spending figures. Our own reading is that scanning is already routine, while evaluation, evidence and automated reporting are harder to do well. If that is where the scarce work sits, a budget signer should look there first, and at who is accountable for it. Endor Labs also says every bad package blocked before it enters is one nobody has to assess or report.
FedRAMP called this "only the first of many migrations", per Endor Labs. The old paperwork was a record of effort. The new record is one of reasoning, and reasoning has to be kept.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Endor Labs.





