---
title: Information Security Risk Register Template (With AI Prompts)
url: "https://goremy.ai/blog/articles/information-security-risk-register-template-ai-risk-scoring"
author: Luis Chavez-Mattos
editor: Luis Chavez-Mattos
editor_title: Director, Product
published: "2026-11-22T08:00:25.087Z"
updated: "2026-11-22T08:00:25.087Z"
topic: Contracts and Compliance
description: Copy a ready-to-use information security risk register template, see a filled-in example, and use AI prompts to score risks and draft treatment plans.
---

# Information Security Risk Register Template (With AI Prompts)

> Copy a ready-to-use information security risk register template, see a filled-in example, and use AI prompts to score risks and draft treatment plans.

Edited by Luis Chavez-Mattos, Director, Product · Published 2026-11-22

## What is an information security risk register template?

An information security risk register template is a table you fill in to record every security risk your organization has identified, how likely and how damaging each one is, who owns it, what controls reduce it, and what you plan to do next. One row is one risk. The register is the single place where security, legal, compliance and business owners agree on what could go wrong and what is being done about it. The template below is ready to copy into Excel or Google Sheets.

## TL;DR

- A good register has **one row per risk, one named owner per row**, and a review date. Rows without owners do not get fixed.
- Score every risk twice: **inherent** (before controls) and **residual** (after controls). The gap shows whether your controls do anything.
- Use a **1 to 5 scale for likelihood and impact** and multiply them. Define each level in writing so two people score the same risk the same way.
- Write risks as **cause, event, consequence** sentences. "Ransomware" is a threat. "A phished admin account lets attackers encrypt the file server and stop invoicing for days" is a risk.
- AI is useful for **drafting risk statements, challenging scores, and proposing treatment options**. It does not know your environment, so a human owner must set the final score.
- Never paste real vulnerabilities, customer data, or credentials into a public AI tool. Describe the risk in general terms or use an approved internal tool.
- A register is only worth keeping if it is **reviewed on a schedule** and changes when something changes.

## What columns should a risk register template include?

Keep it short enough that people fill it in. These columns cover what auditors, risk committees and security leads ask for. Frameworks such as ISO 27001 and the NIST Cybersecurity Framework expect you to identify, assess, treat and monitor risk, and this layout supports all four.

Copy this header row into a spreadsheet:

| Column | What goes in it |
|---|---|
| Risk ID | Unique code, for example SEC-001 |
| Date raised | When it was first logged |
| Risk title | Five to ten words, plain language |
| Risk statement | "Because of [cause], [event] could happen, resulting in [consequence]." |
| Asset or process | The system, data set or business process affected |
| Category | Access, third party, data protection, resilience, application, people, physical, and so on |
| Risk owner | One named person with authority to accept or fund the fix |
| Existing controls | What is already in place and working |
| Inherent likelihood (1-5) | Before controls |
| Inherent impact (1-5) | Before controls |
| Inherent score | Likelihood x impact |
| Residual likelihood (1-5) | With current controls |
| Residual impact (1-5) | With current controls |
| Residual score | Likelihood x impact |
| Rating | Low, Medium, High or Critical, from the score bands |
| Treatment | Mitigate, transfer, avoid or accept |
| Treatment actions | Specific steps, not intentions |
| Action owner | Who does the work |
| Due date | A real date |
| Status | Open, In progress, Overdue, Closed, Accepted |
| Control reference | Link to a policy, framework control or evidence location |
| Next review date | When the owner must look at it again |
| Notes | Decisions, exceptions, links to tickets |

If you are starting from nothing, you can drop to ten columns: ID, title, statement, owner, existing controls, residual likelihood, residual impact, score, treatment actions, next review. Add the rest when someone asks for them.

## How do you score risks in the register?

Scoring fails when "high" means something different to every person. Fix that by writing down what each number means before anyone scores anything.

**Likelihood scale (over the next 12 months)**

| Score | Label | Meaning |
|---|---|---|
| 1 | Rare | Not expected to happen; no history of it here or in similar organizations |
| 2 | Unlikely | Could happen, but would take unusual circumstances |
| 3 | Possible | Has happened in similar organizations; could happen here |
| 4 | Likely | Has happened here before, or controls are known to be weak |
| 5 | Almost certain | Expected to happen, or is already happening |

**Impact scale**

| Score | Label | Meaning (adapt thresholds to your organization) |
|---|---|---|
| 1 | Minor | Absorbed in normal operations; no reportable event |
| 2 | Moderate | Short disruption to one team; limited data exposed; no regulator notice |
| 3 | Significant | Multi-day disruption to a business process; some customer impact; possible contractual breach |
| 4 | Major | Regulator notification likely; material financial loss; lasting customer or press attention |
| 5 | Severe | Threat to the business; large-scale data breach; sustained outage of critical services |

Fill in dollar amounts and record counts that fit your company. Finance and Legal should agree to them once, so the numbers stay consistent.

**Rating bands (likelihood x impact)**

| Score | Rating | Typical expectation |
|---|---|---|
| 1 to 4 | Low | Monitor; accept if the owner agrees |
| 5 to 9 | Medium | Treat within the planning cycle |
| 10 to 16 | High | Treatment plan with a date; report to the risk committee |
| 17 to 25 | Critical | Escalate now; senior sign-off required to leave open |

Some teams build a 5x5 heat map from these bands. The bands are a convention, not a standard, so change them if your risk appetite statement says otherwise. If your company has an enterprise risk framework, use its scales so security risks can be compared with other business risks.

## What does a filled-in information security risk register look like?

Here are three example rows. They are illustrative and not tied to any real company.

| ID | Risk statement | Owner | Existing controls | Inherent L x I | Residual L x I | Treatment | Actions | Due |
|---|---|---|---|---|---|---|---|---|
| SEC-001 | Because administrator accounts do not all require multi-factor authentication, a phished password could give an attacker control of core systems, causing a multi-day outage and data exposure. | Head of IT Operations | MFA on email and VPN; phishing training twice a year | 4 x 5 = 20 | 3 x 5 = 15 (High) | Mitigate | Enforce MFA on all admin accounts; remove unused admin accounts; review admin list quarterly | Q2 end |
| SEC-002 | Because a payroll software vendor holds employee personal data and has not been assessed in two years, a breach at the vendor could expose employee records and trigger notification duties. | Director of HR Operations | Contract has breach notification clause | 3 x 4 = 12 | 3 x 3 = 9 (Medium) | Mitigate / transfer | Collect current security report; confirm breach clause timelines; check cyber insurance covers vendor incidents | Next quarter |
| SEC-003 | Because backups are stored in the same environment as production and restores are not tested, ransomware could destroy both, stopping invoicing for an extended period. | CIO | Nightly backups | 3 x 5 = 15 | 3 x 5 = 15 (High) | Mitigate | Add offline or immutable backup copy; run a documented restore test; record recovery time | 90 days |

Notice SEC-003. Residual equals inherent, which means the existing control is not reducing the risk. That is the kind of finding a register should surface. Also notice that each statement names a cause, an event and a consequence, and that each action can be checked as done or not done.

## How do you build a risk register from scratch?

1. **Set the scales first.** Agree on the likelihood scale, impact scale and rating bands above with Security, Legal, Compliance and Finance. Write them on the first tab of the workbook.
2. **List your assets and processes.** Start from what the business depends on: customer data, payment systems, the HR system, key vendors, source code, laptops. Ten to twenty is enough to start.
3. **Gather risks from several sources.** Pull from the last audit findings, penetration test reports, incident records, vendor assessments, and a short conversation with each department head. Ask "what would stop you working for a week?"
4. **Write each risk as a statement.** Cause, event, consequence. One risk per row. If a row contains the word "and" three times, split it.
5. **Assign one owner per risk.** The owner is the person who can approve the budget or the exception, not necessarily the person who does the technical work.
6. **Record existing controls honestly.** List only controls that are in place and operating. A policy that nobody follows is not a control.
7. **Score inherent, then residual.** Do it with the owner in the room. Challenge scores that cluster at "3 x 3".
8. **Pick a treatment.** Mitigate (reduce it), transfer (insurance or contract), avoid (stop the activity), or accept (documented decision by the owner or a senior approver).
9. **Set actions and dates.** An action without a due date is a wish.
10. **Schedule reviews.** High and Critical risks monthly or quarterly, Medium twice a year, Low yearly. Also review after any incident, major system change or new vendor.

## How can AI help score and write risk register entries?

AI does well at three jobs here: turning a vague worry into a well-formed risk statement, offering a first-pass score with reasoning you can argue with, and suggesting treatment options you may not have considered. It does badly at knowing what is true in your environment. Treat every output as a draft from a smart colleague who has never seen your network.

Before you paste anything, remove names, hostnames, IP addresses, customer details and anything that describes an unpatched weakness in a way an attacker could use. Use your company's approved AI tool if you have one.

### Prompt 1: Turn a worry into a risk statement

```
You are an information security risk analyst. Rewrite the concern below as a risk register entry.

Concern: [paste a plain-language description, no sensitive details]

Return:
1. A risk title (under 10 words)
2. A risk statement in the form "Because of [cause], [event] could happen, resulting in [consequence]."
3. The asset or process affected
4. A category (access, third party, data protection, resilience, application, people, physical)
5. Three questions I should answer before scoring this risk

Do not invent facts about my environment. If something is unknown, list it as a question.
```

### Prompt 2: Get a challengeable first-pass score

```
Act as a risk reviewer. Score the risk below using these scales.

Likelihood (next 12 months): 1 Rare, 2 Unlikely, 3 Possible, 4 Likely, 5 Almost certain.
Impact: 1 Minor, 2 Moderate (short disruption to one team, no regulator notice), 3 Significant (multi-day disruption of a business process, possible contract breach), 4 Major (regulator notification likely, material financial loss), 5 Severe (threat to the business).

Risk statement: [paste]
Existing controls: [paste, or "none known"]
Context: [industry, company size, regulated data types, in general terms]

Give me:
- Inherent likelihood and impact, with a one-sentence reason each
- Residual likelihood and impact given the controls, with a reason each
- The assumptions you made
- Evidence that would change your score up or down

Be explicit that this is a draft for the risk owner to confirm.
```

Check the answer for assumptions you know are wrong, such as a control the model assumed exists. Ask the owner to agree or change each number. If AI and the owner disagree, the reasons are worth recording in the Notes column.

### Prompt 3: Propose treatment options

```
For the risk below, propose treatment options under four headings: mitigate, transfer, avoid, accept.

Risk statement: [paste]
Existing controls: [paste]
Constraints: [budget, team size, systems we cannot change]

For each mitigation, state the action in one sentence, who would likely do it (role, not name), rough effort (days, weeks, months), and how we would verify it worked. Flag any option that depends on something I have not told you.
```

Pick from the options, then rewrite the actions with real owners and dates. AI will not know that your identity team is mid-migration or that the budget closes in March.

### Prompt 4: Audit the whole register for quality

Paste a sanitized export of the register (IDs, statements, scores, owners and dates only).

```
Review this risk register for quality. Identify:
1. Risks with no owner, no due date, or actions that are vague
2. Risks where residual score equals inherent score (controls may not be working)
3. Risks that look like duplicates or should be split
4. Scores that look inconsistent with similar risks
5. Obvious gaps: common risk areas that no row covers, given the categories present

Present findings as a table: Risk ID, issue, suggested fix. Do not rewrite the register.
```

This catches sloppy rows quickly. The gap check is a prompt for discussion, not a verdict on completeness.

### Prompt 5: Summarize for a risk committee

```
Write a one-page summary for a risk committee from the register data below. Include: the number of risks by rating, the top five by residual score with one line each, risks past their due date, and three decisions we need from the committee. Use plain language. Do not add risks or numbers that are not in the data.

[paste sanitized data]
```

Check every number in the summary against the sheet. Models miscount.

## Where should you not trust AI in a risk register?

- **Final scores.** Likelihood depends on your controls, your history and your people. A model can reason about it, but it cannot observe it.
- **Control effectiveness.** AI will assume that a control named in your description works. Only testing or evidence shows that.
- **Legal and regulatory conclusions.** Whether an event triggers a notification duty is a question for Legal. Use AI to prepare questions, not to answer them.
- **Completeness.** A register an AI helped write can look thorough and still miss the one risk specific to your business.
- **Sensitive details.** Do not paste real vulnerability findings, credentials, architecture diagrams, or personal data into a tool your company has not approved.
- **Accepting risk.** Acceptance is a human decision with a name and a date attached.

## How often should a risk register be reviewed?

Set review dates by rating and keep to them. A common pattern is monthly or quarterly for High and Critical risks, twice a year for Medium, and yearly for Low. Trigger an extra review after a security incident, a major system change, a merger, a new critical vendor, or an audit finding. At each review, the owner confirms the description is still true, re-scores likelihood and impact, checks whether actions are on track, and closes or accepts the risk if appropriate.

Keep closed risks in the register, marked closed, with the date and evidence. Auditors ask to see how risks were resolved, and history helps you notice the same risk coming back.

## What mistakes make risk registers useless?

- **Threat lists posing as risks.** "Phishing, malware, insider threat" tells nobody what to do.
- **Everything scored 3 x 3.** If every row lands in the middle, nobody has made a judgment.
- **No owner or a team as owner.** "IT" cannot accept a risk. A person can.
- **Only inherent scores.** Without residual scores you cannot show that controls reduce anything.
- **Actions that are intentions.** "Improve vendor oversight" is not an action. "Collect and review the current security report from the top ten vendors by data volume" is.
- **A register nobody opens.** If it is only updated before an audit, it records the audit, not the risk.
- **A spreadsheet that gets copied.** Five versions in five inboxes means no version is current.

## Can you run a risk register without a spreadsheet?

You can, and many teams eventually want to. Spreadsheets break down when several owners need to update their own rows, when you need reminders for overdue actions, when you want an audit trail of who changed a score, or when only some people should see sensitive rows. Dedicated governance, risk and compliance tools solve this but often cost a lot and come with workflows you never asked for. A middle path is a small internal app shaped around your own columns and approval steps.

## What could a custom risk register app look like with Remy?

If your register has outgrown a spreadsheet, you can describe it in plain English to [Remy](https://goremy.ai) and have it build a small web app around the columns above: owners log in and update only their own risks, the score and rating calculate automatically from your scales, overdue actions trigger email reminders, and Security and Compliance get a dashboard of risks by rating. Remy builds from a written spec and deploys the app with sign-in, a database and audit logging included, and you own the code. Your scoring scales and approval rules stay yours to define.

## Frequently Asked Questions

### What is the difference between a risk register and a risk assessment?

A risk assessment is the activity of identifying, analyzing and evaluating risks. The risk register is the record of the results. You run assessments repeatedly, and the register is the living document they feed. Auditors usually want to see both: the method you used and the register it produced.

### How many risks should be in an information security risk register?

There is no correct number. Start by listing ten to twenty assets and processes, then write the risks that come from them one per row. Focus on whether each risk has an owner, a score and a current review date.

### Does an ISO 27001 risk register look different?

The structure is largely the same. ISO 27001 expects a defined risk assessment method, documented results, and a risk treatment plan that links to the controls you select, which feed your Statement of Applicability. Add a column that maps each treatment to a control reference, and check the control numbering against the current version of the standard you are certified to.

### Should accepted risks stay on the register?

Yes. Keep them with the name of the person who accepted the risk, the date, the reason, and a date when the acceptance expires. Acceptance should not be permanent, because the threat, the business and the controls all change.

### Can I use AI to fill in the whole register?

You can use it to draft rows and challenge your thinking, but not to replace the people who own the risks. AI cannot see your controls, your incident history or your budget. Use it for first drafts of statements, scores to argue about, and treatment ideas, then have each owner confirm the final entry.
