# Risk acceptance vs risk avoidance

Canonical: https://remediationvsmitigation.com/guides/risk-acceptance-vs-risk-avoidance/

Risk acceptance keeps the asset because the risk is within tolerance; risk avoidance removes it because the risk is not (NIST SP 800-39). A BOD 26-02 case mitigates and accepts a real FortiOS flaw on 19 Nov 2026 on a firewall that cannot migrate to a fixed release, then closes it as avoided at replacement on 15 May 2027.

## The line NIST draws

NIST SP 800-39 gives each response its own trigger ([Task 3-1](https://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-39.pdf)). "Risk acceptance is the appropriate risk response when the identified risk is within the organizational risk tolerance." "Risk avoidance may be the appropriate risk response when the identified risk exceeds the organizational risk tolerance." Avoidance "involves taking specific actions to eliminate the activities or technologies that are the basis for the risk". NIST's example ends with eliminating a networked connection and using an "air gap".

For software, SP 800-40 Rev. 4 gives the two in concrete terms ([Section 2.1](https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-40r4.pdf)). Accept: rely on "existing security controls to prevent vulnerability exploitation" or find the impact low enough that no action is needed. Avoid: "uninstalling the vulnerable software, decommissioning assets with the vulnerabilities, or disabling computing capabilities in assets that can function without them."

Acceptance and avoidance compared; SP 800-39 Task 3-1 and SP 800-40 Rev. 4 Sections 2.1 and 2.2, read 19 Nov 2026.

| Question | Risk acceptance | Risk avoidance |
|---|---|---|
| What changes on the asset | Nothing, or only controls around it | The software is removed, the asset retired or the capability turned off |
| Record to keep | Signed [form](https://remediationvsmitigation.com/risk-acceptance-form/) with an expiry; a row in the [exception register](https://remediationvsmitigation.com/exception-register/) | Decommission or removal record; a rescan that no longer finds the asset |
| What NIST says to check afterward | That the response "continues to be in place" | "verify that vulnerable assets were decommissioned or replaced", and that no one "restarts the decommissioned asset" |
| Ends when | The expiry date, a trigger, or a fix | The asset or feature no longer exists |

## Which response fits which situation

Decision table for acceptance and avoidance; sources read 19 Nov 2026. A row gives the response the source supports, not a rule for every organization.

| Situation | Response | Why, and from where |
|---|---|---|
| Risk is inside stated tolerance, no directive applies | Accept, with expiry | SP 800-39 Task 3-1: acceptance is appropriate within tolerance |
| Risk is above tolerance and the feature or software is not needed | Avoid: uninstall or disable | SP 800-40 Rev. 4 lists disabling "computing capabilities in assets that can function without them" under avoid |
| Risk is above tolerance, the asset is needed, the fix is weeks away | Mitigate now, accept the rest for a dated period, fix on a milestone | SP 800-39 describes accepting "for a limited time", then avoiding or mitigating in sequence |
| A federal civilian agency holds an end-of-support edge device | Avoid by the directive's dates | BOD 26-02 requires decommissioning; see the case below |
| A federal civilian agency cannot meet a catalog due date | Remove the asset or apply the catalog action; no acceptance | CISA "does not issue waivers or exceptions for actions required in cyber directives" ([guidance](https://www.cisa.gov/news-events/directives/bod-26-04-implementation-guidance-prioritizing-security-updates-based-risk)) |
| A critical mission system has no newer technology or is too isolated to replace | Accept narrowly, with controls and a replacement plan | SA-22 guidance names these as exceptions to replacing unsupported components ([SP 800-53 Rev. 5](https://github.com/usnistgov/oscal-content/blob/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json)) |

SP 800-39 treats the responses as a sequence, not a menu: in an emergency an organization "might accept the risk associated with unfiltered connection to an external communications provider for a limited time; then avoid risk by cutting the connection; mitigate risk in the near-term" and then mitigate in the long term. Accept first, avoid on a date is therefore NIST's own pattern.

## Worked case: a firewall that cannot migrate, replaced under BOD 26-02

BOD 26-02 times its required actions from issuance. At twelve months agencies decommission devices on the CISA EOS Edge Device List and inventory edge devices that will reach end of support within the succeeding twelve months. At eighteen months they "Decommission all identified EOS edge devices from agency networks", and at twenty-four months they decommission devices "on or before the date these devices reach EOS". Each decommission is reported "to CISA using the CISA-provided reporting template" ([BOD 26-02](https://www.cisa.gov/news-events/directives/bod-26-02-mitigating-risk-end-support-edge-devices), Required Actions). The timeline below prints the dates.

The finding is real. Fortinet advisory FG-IR-26-273, published 10 Nov 2026 and updated 16 Nov 2026, covers CVE-2026-126259, a missing authentication flaw in FortiOS that lets an unauthenticated attacker on the high availability (HA) segment run commands as super admin. Fortinet rates it Critical, CVSS 3 score 9.0, and lists "Known Exploited: No". Fixes are FortiOS 8.0.1, 7.6.8, 7.4.13 and 7.2.14. For FortiOS 7.0, all versions, the solution reads "Migrate to a fixed release", so the fix on that branch is a move to another branch, such as 7.2.14, not a patch. The device is invented: a FortiGate pair at a federal civilian agency, running FortiOS 7.0 on an HA link, with a vendor end-of-support date of 30 Jun 2027. By the example's assumption, the pair's hardware model is not supported by the 7.2 line, so migrating to 7.2.14 is closed to it and replacement is its only way to a fixed release. On 19 Nov 2026 it is still vendor-supported, so the directive does not yet require anything of it, and the CVE is not a CISA catalog entry in the 2026.11.18 feed. Fortinet's text says exploitation needs access to the HA segment, so the agency can narrow the exposure without an upgrade.

Timeline of the worked case. Directive dates are from BOD 26-02 and the advisory facts from Fortinet's page for FG-IR-26-273, both read on 19 Nov 2026; the firewall, its vendor end-of-support date, the acceptance, its backstop expiry and the signer are invented.

| Date | Step | Response | Real or invented |
|---|---|---|---|
| 5 Feb 2026 | BOD 26-02 issued | None | Real |
| 19 Nov 2026 | The agency's scan matches F-2210 to CVE-2026-126259 on the firewall. The authorizing official accepts that finding, not the age of the device, until the device is replaced. Controls: the HA link sits on a dedicated segment reachable by no other host, and management access is limited to one jump host. Backstop expiry 31 May 2027 (193 days). Replacement is ordered | Mitigate and accept | Real CVE; invented device and decision |
| 5 Feb 2027 | Twelve-month deadline. The firewall appears in the inventory of edge devices that will become end of support within the succeeding twelve months. Its end-of-support date is after 5 Feb 2027, so it is not yet due for decommissioning | Accept, review | Real date; invented firewall |
| 15 May 2027 | Cutover to the new device, 177 days after the acceptance. Old device powered off and removed from the inventory; a rescan finds nothing at its address; the decommission is reported to CISA on the CISA-provided reporting template (BOD 26-02, Required Actions 3 and 4). The acceptance of F-2210 is closed the same day with the status "avoided", with the decommission record and the rescan attached; the device no longer exists, so nothing remains to accept | Avoid, close | Invented |
| 30 Jun 2027 | Vendor end-of-support date, 46 days after cutover | None | Invented device date |
| 5 Aug 2027 | Eighteen-month limit for every identified end-of-support edge device, 36 days after the vendor date | None | Real |

Two points carry over to any case. First, the acceptance ends at cutover, and its backstop expiry is earlier than the vendor's end-of-support date, so the decision never outlives the device or the support it relies on. If cutover slipped past 31 May 2027, the acceptance would have to be renewed on a new decision, not left open. The firewall is inventoried at twelve months as going end of support, so the twenty-four-month action applies to it: decommission "on or before the date these devices reach EOS", which is 30 Jun 2027 here, earlier than the 5 Aug 2027 eighteen-month limit. Second, a device that cannot be replaced by the directive's date is not an acceptance question for that agency: the agency would contact CISA, and the [BOD 26-04 guidance](https://www.cisa.gov/news-events/directives/bod-26-04-implementation-guidance-prioritizing-security-updates-based-risk) says CISA issues no waivers for directive actions. Outside federal agencies the same sequence works with the organization's own dates, as in [accepting risk on end-of-life software](https://remediationvsmitigation.com/guides/accepting-risk-on-end-of-life-software/) and the [end-of-life row of the Decision Table](https://remediationvsmitigation.com/decision-table/end-of-life-software/).

## The same action under three names

Decommissioning is avoidance in SP 800-40 Rev. 4 and remediation in CISA BOD 26-04, which says a vulnerability "shall be remediated (that is, the vulnerability is eliminated through patching, decommissioning the system, or another action)" ([BOD 26-04](https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk)). A register row for a decommission should therefore record the action, not only the word.

A second overlap sits inside NIST's own list. Disabling "a vulnerable feature" is filed under mitigate, while disabling capabilities in assets that can function without them is filed under avoid. Pick by what remains: if the vulnerable code stays installed but unreachable, record mitigation; if it is gone, record avoidance. The [disable a feature or service control](https://remediationvsmitigation.com/compensating-controls/disable-feature-or-service/) covers the first case, and [risk transfer vs acceptance](https://remediationvsmitigation.com/guides/risk-acceptance-vs-risk-transfer/) the remaining response.

## Sources

- NIST SP 800-39, Task 3-1: https://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-39.pdf (read 19 Nov 2026)
- NIST SP 800-40 Rev. 4, Sections 2.1 and 2.2: https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-40r4.pdf (read 19 Nov 2026)
- CISA BOD 26-04, definitions beneath Table 1: https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk (read 19 Nov 2026)
- CISA BOD 26-04 implementation guidance, waiver question: https://www.cisa.gov/news-events/directives/bod-26-04-implementation-guidance-prioritizing-security-updates-based-risk (read 19 Nov 2026)
- CISA BOD 26-02, Required Actions: https://www.cisa.gov/news-events/directives/bod-26-02-mitigating-risk-end-support-edge-devices (read 19 Nov 2026)
- NIST SP 800-53 Rev. 5 catalog (OSCAL), SA-22 guidance: https://github.com/usnistgov/oscal-content/blob/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json (read 19 Nov 2026)
- Fortinet PSIRT advisory FG-IR-26-273 (CVE-2026-126259), published 10 Nov 2026, updated 16 Nov 2026: https://fortiguard.fortinet.com/psirt/FG-IR-26-273 (read 19 Nov 2026)

Last reviewed 19 Nov 2026

## Related

- [Risk Acceptance Criteria: 7 Tests From NIST and CISA](https://remediationvsmitigation.com/guides/risk-acceptance-criteria/index.md)
- [End-of-Life Software: Replace, Isolate or Accept With Dates](https://remediationvsmitigation.com/decision-table/end-of-life-software/index.md)
- [Risk Acceptance Form: Free Template With 15 Fields](https://remediationvsmitigation.com/risk-acceptance-form/index.md)
- [Exception Register: 15 Columns, States and Blank CSV](https://remediationvsmitigation.com/exception-register/index.md)
- [Risk Transfer vs Risk Acceptance for Security Findings](https://remediationvsmitigation.com/guides/risk-acceptance-vs-risk-transfer/index.md)
- [Risk Acceptance vs Risk Tolerance: Who Sets the Limit](https://remediationvsmitigation.com/guides/risk-acceptance-vs-risk-tolerance/index.md)
- [Accepting Risk on End-of-Life Software: Limits and Record](https://remediationvsmitigation.com/guides/accepting-risk-on-end-of-life-software/index.md)
