SaaS or a managed service you cannot patch

When the vendor owns the fix, change what you control and track the vendor's date. If the vendor has already patched, close the entry on its written evidence and check your tenant for earlier misuse. FedRAMP's 192-day acceptance limit and the 7 Dec 2026 FedRAMP deadline set the outer dates.

  1. Remediate
  2. Mitigate now, fix laterGuidance points here
  3. Accept with a dated exceptionAlso allowed

Signals: Vendor controls the fix; Your lever is configuration, access or exposure

Who owes what

Quotes read on 19 Nov 2026 from FedRAMP's agency POA&M page and CISA's BOD 26-04 implementation guidance.
PartyDutySource and quote
Cloud providerDetect, evaluate, report, mitigate and remediateFedRAMP: "A FedRAMP Certified cloud service provider is responsible for detecting, evaluating, reporting, mitigating, and remediating vulnerabilities in its cloud service offering."
Customer (agency)Write its own plan only for actions it ownsFedRAMP: agencies create POA&Ms "only when there are actions the agency needs to take, track, fund, manage, or accept"
Customer (agency)Stay engaged with providersCISA guidance: agencies "remain responsible for engaging service providers directly, as needed, to ensure compliance with CISA Directives."
Commercial providerReport outstanding vulnerabilities to the customerCISA guidance: the provider "must report the status of outstanding vulnerabilities to the agency receiving the service"

NIST SP 800-40 Rev. 4 puts SaaS under Transfer: "replacing conventional software installations with software-as-a-service (SaaS) usage where the SaaS vendor/managed service provider takes care of patching." Its definition of Transfer is "Reduce the risk by sharing some of the consequences with another party". The exposure is still there until the vendor fixes it, so while a vendor fix is expected the answer is mitigation and a tracked fix, not acceptance. Acceptance comes in only when the vendor says it will not fix the flaw, or the fix passes the 192-day limit below. The customer-side controls for tenants are listed in SaaS vulnerability mitigation; this row decides which response to record and when to close it.

A hosted service patched before the customer knew: BeyondTrust, 29 Aug to 16 Oct 2026

CVE-2026-102764 is an origin-validation flaw in BeyondTrust Remote Support that lets an unauthenticated remote attacker perform file operations on the appliance. BeyondTrust patched its hosted (SaaS) customers itself, so no hosted customer could patch at any point. The advisory dates the SaaS patch as of 9 Oct and the blog dates the patch release 8 Oct. Day counts are from 29 Aug.

Sources: BeyondTrust BT26-05 and blog, CISA KEV feed 2026.11.18, all read 19 Nov 2026.
DateEventDayWhat a hosted customer could do
29 Aug 2026A suspected China-nexus actor exploits the flaw against one BeyondTrust-hosted instance and one self-hosted appliance0Nothing: no notice existed
28 Sep 2026BeyondTrust threat hunting finds the hosted-instance intrusion and contains it within 27 minutes30Nothing: no notice existed
8 Oct 2026BeyondTrust releases the patch; SaaS instances start to be patched automatically (blog)40Nothing to install
9 Oct 2026Advisory: "As of 9 Oct 2026, a patch that remediates this vulnerability has been applied to all Remote Support SaaS customers"41Nothing to install
15 Oct 2026Coordinated disclosure: CVE, advisory BT26-05 (CVSS v4 9.3), blog post, Mandiant analysis47Decide the register entry
16 Oct 2026CISA adds the CVE to KEV with a due date of 19 Oct 2026, 3 days later48Federal agencies: follow the BOD 26-04 cloud guidance

On 15 Oct a hosted customer faces a choice between two records. The vulnerability entry can close at once: the advisory confirms the SaaS patch as of 9 Oct, and the blog says "BeyondTrust SaaS customers have already been patched, and no further action is required." Keep the advisory page and its 15 Oct issue date as the closing evidence. No mitigation entry is needed for a flaw that is already fixed.

The second record is not closed by the patch. The flaw was exploited for 40 days before the patch was released and 41 before the advisory confirms it on every SaaS instance, and BeyondTrust writes that it is "only aware of the two impacted customers" and has "not seen evidence of adversary activity since August". That is the vendor's statement about its own telemetry. Open an entry in the incident or security-review log (it is a check for past misuse, not a control), ask the provider in writing whether your tenant was among the affected instances, and close it with the reply and the review below.

What a customer can still review in its own tenant

BeyondTrust's blog lists what to hunt for after this actor. Four of its items can be read in a hosted tenant's own audit and session records, and each is a customer-side check:

  • Privileged or administrative sessions from residential broadband or commercial VPN addresses that nobody can tie to a known administrator or a ticket.
  • Unauthorized account, permission or session-policy changes, "particularly modified permissions for 'Jump Items' or 'Teams'".
  • New multi-factor registrations on privileged accounts.
  • Stored credentials authenticating outside a managed session or from unfamiliar locations. If anything is found, BeyondTrust says: "Rotate every credential the appliance stored or injected."

Record the review in that incident or review log with the date it was run and the period covered, 29 Aug to 9 Oct 2026. It does not belong among the compensating controls, which stand in for a fix. For a flaw with no vendor fix yet, controls do apply, and those go in that register: switch off the affected feature, limit which networks and accounts reach the tenant, and rotate any credential the advisory names. FedRAMP's rules describe the provider's duty in the same terms: "If it is not possible to fully mitigate vulnerabilities or remediate vulnerabilities, providers SHOULD instead partially mitigate vulnerabilities promptly, progressively, and persistently."

When the vendor has not fixed it

FedRAMP's rule VER-TFR-MAV says providers "MUST categorize any vulnerability that is not or will not be fully mitigated or remediated within 192 days of evaluation as an accepted vulnerability." Counted from an evaluation on 15 Oct 2026, 192 days ends on 25 Apr 2027 (our count). CISA's BOD 26-04 guidance says FedRAMP "will require mandatory adoption of the new Vulnerability Detection and Response (VDR) and Vulnerability Evaluation and Reporting (VER) rules by 7 Dec 2026". The older FedRAMP playbook, reference only since 24 Jun 2026, asked for high-risk vendor dependencies to be "mitigated to a Moderate level through compensating controls within thirty (30) days" and for a vendor check-in "at least once a month". No source sets a number for non-federal customers, so set your own review date from the provider's committed fix date and write the reason beside it.

Record to keep, review date, what changes the answer

  • Record: the provider's advisory or status notice with its date, your configuration changes and the dates they were tested, the contact who confirmed the vendor's fix plan, and the date of the last vendor check-in.
  • Review date: the provider's committed fix date, or your own date with the reason written down. For a vendor with no fix date, review monthly.
  • Changes the answer: the provider publishes the fix (record the close date); the flaw reaches the CISA catalog (the provider's due date then applies, see patch available and exposed); or the provider will not fix it, or the 192-day limit passes, which turns the entry into a dated exception or an exit plan. That is the only point at which acceptance is the right record.

The limits of transfer are covered in risk transfer versus risk acceptance, and the plan format for agencies in the POA&M guide.

Sources

Last reviewed