3PAOs like Fortreum have accepted reachability analysis as evidence for FedRAMP false positive determinations since 2024. What was missing was the rulebook.
The FedRAMP Consolidated Rules for 2026, mandated through the FedRAMP Program Management Office (PMO) under GSA and effective for optional adoption on July 4, 2026, close that gap: cloud service providers are now required to evaluate every detected vulnerability for exploitability and reachability before deciding what to fix.
What auditors endorsed case by case is now formalized as a program-wide requirement.
What changed in the 2026 rules
The Vulnerability Evaluation and Reporting ruleset replaces the old model, where every scanner finding started life as a POA&M line item with a CVSS-driven remediation clock, with an evaluation-first model. Three requirements matter most:
Providers must evaluate exploitability (VER-EVA-ELX)
A likely exploitable vulnerability (LEV) is one where, given the context of your cloud service offering, exploitation is realistically achievable during normal operation rather than theoretically possible.
Every detected vulnerability must be assessed against that bar, in the context of your offering. FedRAMP's own guidance note is blunt about the likely outcome: most traditional vulnerabilities discovered by scanners are not likely to be exploitable, because exploitation typically requires an unrealistic set of circumstances.
Providers must evaluate internet-reachability (VER-EVA-EIR)
An internet-reachable vulnerability (IRV) is one that can be triggered by processing a payload that originated on the internet, even if the vulnerable system is never directly exposed.
FedRAMP deliberately chose "internet-reachable" over "internet-accessible." A service behind a load balancer and firewall can still process a malicious payload that originated on the internet. Log4Shell is their canonical example: the vulnerable resource sat deep in the application stack, but attacker-controlled data reached it anyway.
Providers should evaluate false positives (VER-EVA-EFP)
The evaluation factors providers must consider now name reachability and exploitability as factors two and three, right after criticality. Reachability analysis went from a methodology you had to defend to a factor FedRAMP tells you to use.
The classification you assign drives everything downstream: reporting obligations, incident thresholds, and timelines. Internet-reachable LEVs with high Potential Agency Impact ratings can escalate to FedRAMP Reportable Incidents. Vulnerabilities that fail the exploitability test don't.
A long time coming
None of this should surprise anyone who's worked with a forward-leaning 3PAO. Back in 2024, Fortreum's Ben Scudera published an auditor's analysis of FedRAMP vulnerability management costs after a client asked Fortreum to review Endor Labs' reachability analysis as a means of automatically identifying false positives. Fortreum agreed the concept had merit and presented it to FedRAMP for approval.
Scudera's bar for a defensible false positive determination was specific. The tool must distinguish "we can't determine reachability" from "we're confident this is not reachable." An unreachable determination must hold from any source, not just the public internet. And the tool must alert you when a code change makes an unreachable vulnerability reachable.
Those three criteria map almost exactly onto what the 2026 rules now demand of every provider. The difference is that evaluation is no longer optional, and the timelines are tight: depending on your certification class, you're expected to complete evaluation within 2 to 14 days of detection. At the volume a typical CSO's scanners produce, that's not a job for spreadsheets and engineering judgment calls. It requires automated, evidence-backed analysis.
And it's a double shift: traditional ConMon reporting was measured in months, and the analysis behind it was surface-level. The 2026 rules compress the timeline and deepen the analysis at the same time, both variables changing at once.
There's also a supply-side reason the old model was breaking either way. CVE submissions grew 263% between 2020 and 2025, and Q1 2026 is running nearly a third higher than Q1 2025. In our own vulnerability database, the first 100 days of 2026 produced roughly as many new open source vulnerabilities as all of 2025 combined.
NIST responded this April by limiting enrichment to prioritized CVEs. Everything else is now marked "Not Scheduled" and arrives with no CVSS score at all. A remediation clock keyed to CVSS doesn't work when most vulnerabilities never get one. Evaluation-first is what's left.
Internet reachability comes first
An internet-reachable vulnerability, by FedRAMP's definition, is one where a payload from the internet can be processed by the vulnerable resource. Of course, if your application never calls the vulnerable function, there is no execution path for any payload to trigger it, from the internet or anywhere else. A vulnerability that isn't reachable in code cannot be internet-reachable.
That gives you a clean, layered evaluation model:
Layer 1: Function-level reachability eliminates the majority of findings
Endor Labs builds a call graph of your application and determines whether any path exists from your code to the vulnerable function. Our research shows fewer than 9.5% of vulnerabilities are exploitable at the function level. Everything unreachable is strong, auditable evidence for a "not likely exploitable" determination and a false positive classification, exactly the evidence VER-EVA-EFA factor two asks for. This analysis has already been evaluated by the FedRAMP PMO and 3PAOs including Fortreum and Schellman and meets their standard for false positive identification.
When Egnyte pursued FedRAMP authorization, function-level reachability showed that only 9% of its findings were reachable, a 91% noise reduction that cut mean time to remediation by 70% and let a lean product security team run ConMon across more than 350 engineers without adding headcount. As Egnyte's VP of Security Dawid Balut put it, ConMon at FedRAMP scale doesn't work if every finding requires manual security judgment, reachability evidence is what lets developers act on findings without a security engineer in the loop.
Layer 2: Runtime context resolves internet-reachability for what remains
Code-reachable doesn't automatically mean internet-reachable. A vulnerable function your application calls might only process trusted internal data. For that remaining subset, runtime and cloud context from sensors like Microsoft Defender, Upwind, Wiz, and other CNAPP integrations adds the exposure evidence you need to determine whether attacker-controlled payloads can actually arrive, and to assign an honest Potential Agency Impact rating. You can combine Endor Labs’ function-level reachability from static code analysis with container reachability with any of those CNAPP providers.
Endor Labs answers the code-reachability question. Runtime and network context from these partners answers whether an exploit originating outside your system, from your network environment or beyond, can actually reach it.
Static analysis does the heavy lifting on volume. Runtime context does the precision work on what's left. Neither alone answers both questions the 2026 rules ask.
The reporting burden is also an automation problem
Evaluation is half the story. The 2026 rules require monthly human-readable activity reports, machine-readable JSON reporting against FedRAMP schemas, per-vulnerability detail including internet-reachability and exploitability determinations, and reassessment discipline: any vulnerability not fully mitigated within 192 days must be categorized as an accepted vulnerability with a documented justification.
This is where grouping and deduplication matter as much as reachability. Grouping vulnerabilities was a gray area under the old rules; the VER rules now explicitly permit, and encourage, logically grouping them. Deduplication works the same way: if your SCA scan and your container scan find the same log4j instance, that's one finding, not two.
Endor Labs correlates findings across scan types automatically and generates the SBOM and VEX documents that back up every determination, so your monthly report reflects real unique risk instead of scanner overlap.
Rubrik runs its FedRAMP environment on exactly this combination. With near-zero tolerance for vulnerabilities and internal remediation SLAs as short as 24 hours, Head of Security Marty Garvin credits reachability analysis and consolidated findings with shrinking the set of true positives that require remediation, and the team consistently hits those SLAs because nobody has to argue over whether a finding is real.
The determination is the easy part
Most reachability conversations stop at the determination: can you defend a "not likely exploitable" call? But under the 2026 rules, that call is just an input. The deliverable is a per-vulnerability record, in a FedRAMP-defined JSON schema, that an agency reviewer can process without calling you.
VER-RPT-VDT spells out what every detected vulnerability has to carry, roughly a dozen fields, from detection time and reachability determinations through PAIN reductions to final disposition.
That's a data problem as much as a compliance problem. Half the fields come from your scanning and analysis tooling. The other half come from your compliance program over time, as ownership changes and remediation progresses. Neither half is useful alone, and FedRAMP doesn't care that they live in different systems.
The same split shows up when you accept a vulnerability instead of fixing it. VER-RPT-AVI governs accepted vulnerabilities, and VER-TFR-MAV says anything not fully mitigated or remediated within 192 days of evaluation becomes one, with a written explanation of why. That's not a deadline you hit once, it's a clock that starts on every finding at the moment of evaluation, and something has to be watching all of them.
Then there's cadence. VER-TFR-MHR requires a human-readable activity report at least monthly. VER-TFR-MRH asks for machine-readable historical activity available for automated retrieval, refreshed monthly, every 14 days, or every 7 days depending on your certification class. A reachability determination that lives in a security tool and gets exported by hand before each report is a determination you'll eventually get wrong.
Conclusion
FedRAMP spent a decade treating every CVSS 7.0 as a 30-day fire drill regardless of whether the vulnerable code could ever execute. The 2026 Consolidated Rules replace that with a model auditors like Fortreum validated years ago: evaluate first, prove your reasoning with evidence, and spend remediation effort on vulnerabilities that are actually reachable and exploitable.
If you're a CSP facing the December 2026 dates, the question is no longer whether reachability analysis is acceptable, FedRAMP settled that. The question is whether you can produce the evidence inside a 2-to-14-day evaluation window, and whether it survives the trip into a machine-readable report, a PAIN rating, a 192-day clock, and an audit conversation six months later.
That's a two-system problem, and pretending otherwise is how programs end up back in spreadsheets. Endor Labs produces the evidence, Kovr carries it into the reports agencies expect, and Fortreum validates the methodology connecting them.
See how Endor Labs supports FedRAMP vulnerability evaluation and reporting on our FedRAMP use case page, or book a demo to walk through your ConMon workflow.
What's next?
When you're ready to take the next step in securing your software supply chain, here are 3 ways Endor Labs can help: