Outgrace
OutgraceThe Leadership Magazine, home

A board checklist for cyber risk oversight, tested against NIST's framework

Most boards receive a cyber report. Few can say whether it tells them what they need to oversee. Seven questions a director can put to management before the next meeting.

Most boards now receive a cyber report. Far fewer can say what that report would need to contain before it told them something they did not already know. The test in this piece is practical. A director should be able to open the board pack and answer basic questions about the company's exposure without asking management to translate the jargon. NIST's Cybersecurity Framework exists, according to NIST's own site, to help organisations understand and improve cyber risk management, and its stated aim is reducing that risk for industry, government and other organisations. The site does not tell a board what to ask. As we reviewed it, the page does not mention board oversight at all. The checklist below therefore fills that gap with judgement, built on what the UK and EU cyber agencies publish about the same problem.

Chapter one: 1. Know which services the board is protecting

Start with the question that everything else depends on: what must keep working? The UK's National Cyber Security Centre directs its Cyber Assessment Framework at organisations with a vital role in UK daily life, and it focuses on their essential functions. That focus is the right one for a board. An attack on a company is a headline. An attack on the three services that customers cannot do without is a board matter, and the board should know which services those are before any incident occurs.

What usually goes wrong is a list that exists but has not been looked at for years. Management produces a register of systems, sorted by age or by owner, and the board receives it as an appendix. The register tells the board what the company owns. It does not tell the board what the company cannot lose. A sensible test is to ask management to name the two or three services whose interruption would reach customers within a day, and to show when that list was last reviewed and who signed it off. If the answer is a shrug, the board has found its first gap.

Chapter two: 2. Test the defences, not only the reports

Monitoring is necessary, and the NCSC is clear that it is not sufficient. The agency says monitoring alone cannot detect every threat, and that proactive threat hunting can expose gaps in defences that alerts never show. For a board, this changes what a good report looks like. A dashboard that counts alerts closed in a month is measuring activity. It says little about whether an intruder who got past the first line would be found. Ask whether the company hunts for threats in its own systems, how often, and what the last search found, even if the answer is that it found nothing and the team is not sure why.

For critical organisations, the NCSC has also launched a Cyber Adversary Simulation scheme. It aims to strengthen resilience through real-world testing by trusted providers. A board does not need to run such a test itself. It does need to know whether one has been done, by whom, and what it changed. The test is simple: has the board received a finding from a real test in the past year, and did the board see what management did about it? A report that lists controls installed but no test results shows what the company bought, not what it can withstand.

Chapter three: 3. Put recovery on the board agenda

Most boards discuss prevention at length and recovery for a few minutes. The NCSC publishes recovery resources for organisations to recover from cyber attacks and improve their cyber resilience, which signals how much the agency regards recovery as a discipline of its own. In our experience, recovery is where a board can add the most value, because the decisions are commercial: which customers are told first, how long the company will accept a shutdown, and what it will pay to shorten one.

The test is a number. Ask management for the time it would take to restore the company's most important service from backups, and when that time was last measured rather than estimated. If the figure has never been tested, the board should treat the recovery plan as a hypothesis. Write the figure into the minutes, and ask for it again after the next test. A recovery time that appears only in a plan document has not yet been proved.

Chapter four: 4. Track the software you depend on

Vulnerabilities in widely used software reach boards through the news, usually after a customer or regulator has asked about them. The NCSC's recent alert urged UK organisations to act on a PHP security issue identified by two CVE numbers, CVE-2026-109618 and CVE-2026-109619. The useful lesson is not the specific flaw. It is the chain between the alert and the decision. ENISA's vulnerability database, according to the agency, gives aggregated information on mitigation measures and exploitation status, which is the kind of detail a board needs to judge urgency.

The test is one question: when an advisory names a product the company runs, who tells the board, within what period, and what does the board receive? A good answer names an individual and a time limit, and it says whether the affected systems are essential ones. A poor answer is that the IT team deals with it. Boards do not need to read advisories. They need to know that a process exists, that it has been used, and that the company can say which of its systems would be affected by the next one.

Chapter five: 5. Ask about AI, both as attacker and as tool

The NCSC urges organisations to be ready for the enhanced capabilities of AI-powered attacks. That is a warning about the threat, and it arrives alongside a second question that boards often miss: how the company itself uses AI in its software and security work. ENISA has issued a joint advisory on frontier AI models for code review, dated 12 November 2026, and ran an experiment with CERT-EU on how such models could support security reviews of open-source software. Both show that AI tools are now being used in the work of finding weaknesses.

For a board, the test is about control rather than enthusiasm. Ask whether any AI tool touches the company's source code, its security reviews or its incident response, who approved it, and what human check sits between the tool's output and a change to a live system. Ask also what the company has changed in its own defences since the NCSC's warning. If management cannot name a change, the board has learned that the warning has not yet reached the people who would have to act on it.

Seven questions for the next board meeting

Each question can be answered from documents management already holds. A no, or a silence, marks a place to press.

  • Does the board have a written list of the services that must keep running, reviewed within
  • Has the board received the results of threat hunting or a real-world test in the past year
  • Has the recovery time for the most important service been measured, not only estimated?
  • Is there a named person who tells the board about advisories affecting systems the company
  • Can management say which AI tools touch its code, security reviews or incident response?
  • Has management used a maturity assessment and reported the result to the board?
  • Does one named director hold cyber risk on the board's agenda, with time set aside at each

Outgrace checklist

Where it breaks

The checklist fails in three predictable ways. The first is that the board receives the right questions and a report that answers a different one. Management can show green status on every control and still have no test results, so the board should insist on seeing what was tested, not only what was installed. The second is that the cyber item is handed to a committee with no clear remit, and the full board hears only a summary. The third is that the framework becomes a box-ticking exercise. NIST says its resources overlap and share themes, and its profile templates exist so that an organisation can describe its own position. A profile that no director has read does nothing for the board.

A note on sources. The three agency sites we reviewed, NIST, the NCSC and ENISA, do not provide board-specific cyber oversight guidance, and we did not find CISA guidance for boards among them. The checklist therefore rests on what those agencies say about threats, testing, recovery, vulnerabilities and AI, and on our judgement about how a board should ask about them. A board that wants a source written for directors should look for one and compare it with this list before adopting either.

Start on Monday with the first two questions, and ask management to answer them in writing before the next meeting. Ask for the list of essential services, with the date it was last reviewed and the name of the person who approved it. Then ask for the most recent threat hunting or test result, and what changed as a result. A written answer is more useful than a presentation, because it shows what management actually knows. Once the first two answers are in the pack, add the remaining five at the next meeting, and keep the list in the board's minutes so that the answers can be compared year on year.

Sources

Checked against 3 sources

  • NIST (opens in a new tab)Primary source — The framework's stated aim of helping organisations understand and improve cyber risk management, its offer of templates for CSF profiles, and the note that its resources overlap.
  • NCSC (opens in a new tab)Primary source — Guidance on monitoring limits and threat hunting, the Cyber Adversary Simulation scheme, the Cyber Assessment Framework's focus on essential functions, recovery resources, the PHP alert and AI-enabled
  • ENISA (opens in a new tab)Primary source — The joint advisory on frontier AI models for code review, the CERT-EU experiment, the maturity assessment tool, and the vulnerability database's information on mitigation and exploitation status.

In the sectors

Written by Outgrace’s editors with AI-assisted research and drafting, and checked against the sources above. Found something wrong? Our corrections policy explains how to tell us. How we work.

Newsletter

The Briefing

One email a week: the leaders we profiled, the ideas worth stealing, and nothing else.

Weekly, free, one click to leave. Read our privacy policy.