Skip to content
FortaRisks
Back to blogGuidance

Risk register: controls that hold

November 22, 2026 · 6 min read

Your Sunday briefing · Risk series

Last Sunday, four small risks on the same vendor became one big one. Your register now says what can happen, to whom, and through whom. What remains is the column everyone fills in quickly: controls. "MFA on external access." "Daily backups." "Vendor access reviewed." Each entry lowers residual risk by a notch. The leadership team sees green and moves on.

Our position: a control that has never been tested should not lower your residual risk. Until it is verified, it is an intention, not a protection. The profession calls this control effectiveness, and has long told apart the control that is designed from the control that works. This episode brings that distinction into the register. It is a position, not an accounting rule: a control deployed at a known coverage can earn partial credit, as long as you write it down.

Why the register overstates its controls

  • We write the policy, not the reality. The "control" column repeats what the policy says. It does not say whether the control covers every system, or whether it still works.
  • Coverage is never 100%. "MFA everywhere" often means everywhere except the legacy server, the service account, the provider's access. The risk goes precisely through the exception.
  • Nobody has the date of the last test. A control checked three years ago counts as much as one checked last month. The register cannot tell.

What it changes: the question to ask is not "do we have a control?", but "when did we last see it work, and on what?".

What the frameworks say, and what February 2024 showed

NIST, in SP 800-53A on assessing controls (revision 5, January 2022), asks three questions of every control: is it implemented correctly, is it operating as intended, is it producing the desired outcome? Those are three different levels, and most registers stop at the first. The NIST Cybersecurity Framework 2.0 asks that improvements be identified from security tests and exercises, including with suppliers (ID.IM-02). Status: official publications.

The clearest example is the attack on Change Healthcare, a UnitedHealth Group subsidiary, in February 2024. Before the US Senate Finance Committee on May 1, 2024, Andrew Witty, then CEO of UnitedHealth Group, said criminals had used compromised credentials to log into a Citrix remote access portal, and that the portal did not have multi-factor authentication. Answering senators' questions, he added that the group had an MFA policy for its external systems, and that this server, which came with the 2022 acquisition of Change Healthcare, had not yet been brought up to it. Status: confirmed by the company, under oath. On July 31, 2025, the company reported approximately 192.7 million people affected to the Office for Civil Rights of the US Department of Health and Human Services (HHS), a total that may include duplicates. Status: confirmed by HHS.

What it changes: the policy existed. So did the control, almost everywhere. A register would probably have read "MFA on external access: in place". The one thing that mattered was missing: the list of access points actually checked.

The method: from declared control to proven control

1. List the credited controls, scenario by scenario. For each scenario in your register, write the two or three controls that lower its level. Not every control in the company: the ones that matter for this scenario.

2. Give each control a level of evidence. Three levels are enough. Designed: it exists on paper (policy, procedure). In place: it is deployed, with a known coverage ("MFA on 46 of 48 access points"). Effective: it has been tested against the scenario, on a date, with evidence kept.

3. Recalculate residual risk. Our starting rule, to adjust: only an effective control fully reduces the risk; a control in place reduces it in proportion to its coverage; a designed control reduces nothing. Your residual will go up. That is the right number.

4. Test against the scenario, not in the abstract. The test follows from the scenario written in the first episode. For MFA, try a login without a second factor on every external access point on the list. For backups, restore a system, not a file. For vendor access, revoke it and time how long it takes. The test produces evidence: a date, a result, a screenshot.

5. Give the control an owner and a next date. The person who monitors, in the sense of the second episode, holds the date of the next test. A signal (a newly exposed service, a vendor changing hosting provider, an acquisition) sends the control back to "in place" until it is tested again.

The three traps

The forgotten exception. A control at 96% coverage protects 96% of systems, not 96% of the risk. If the scenario goes through the 4%, it protects nothing. Write the exceptions down in full.

Design evidence taken for effectiveness evidence. A signed policy, a configuration screenshot, a point-in-time audit report show that a control was designed. It is the difference between a SOC 2 Type I report, at a date, and a Type II, over a period.

The control that carries ten scenarios. MFA, segmentation or backups appear on half the register. If one fails, ten risks rise at once: test it first, as you would a concentration.

This week's test: thirty minutes, three scenarios

Bring together the person who keeps the register and the person who runs the technical controls.

  1. Take the three most severe scenarios in the register.
  2. For each credited control, ask: date of the last test, on what scope, with what evidence?
  3. Reclassify each control (designed, in place, effective) and recalculate the residual of the three scenarios.

Three results should worry you. No test date on a credited control: the register's green is a hypothesis. "More or less full" coverage: someone knows the exception, but it is written nowhere. A residual that rises by two notches or more: the risk leadership accepted is not the one it carries.

What you should have next Sunday

The same register, with, for the three most severe scenarios, controls classified by level of evidence, the date of the last test, the exceptions written down and a recalculated residual. At least one test scheduled for the control that carries the most scenarios. Your register may show less green: it will finally say what really protects you.

Where FortaRisks fits

The register lives in Forta Cockpit, the decision-makers' view: each risk carries its inherent and residual level, with its treatment and its owner. Forta Actions turns a control that fails its test into an action, with an owner and a due date, tracked until it is resolved. To put your three most severe scenarios through the test with an expert, the prioritization session starts from your assets, your controls and your findings.

Next Sunday, November 29: estimating in ranges. A register says "high". Your board wants to know how much, and an honest range beats a precise, wrong figure. The method is done by hand; in Forta Cockpit, quantification is on our roadmap.

Sources: NIST, SP 800-53A Rev. 5, Assessing Security and Privacy Controls (January 2022) · NIST, Cybersecurity Framework 2.0 (CSWP 29, February 26, 2024) · US Senate Finance Committee, testimony of Andrew Witty (May 1, 2024) · HHS, Office for Civil Rights, Change Healthcare Cybersecurity Incident FAQ

And at your end?

See your attack surface like an attacker, free

Give us your domain. We map your external exposure and hand you a prioritized exposure report within 48 hours, with no agent to install and no obligation.