A working AI risk register for security, risk, and compliance teams. The spreadsheet ships with 30 pre-loaded AI risks across eight categories, 5-point likelihood and impact scoring, automatic inherent and residual ratings, and dropdowns for category and status. Delete what does not apply, add what does, and you have a living register.
Three worksheets, built so your team can start scoring risks the same day.
How to use the register, step by step: tailoring the pre-loaded risks, scoring, naming owners, and keeping the register current.
The working sheet. 14 columns from risk ID to next review date, with formulas that calculate inherent and residual scores as you type.
Shared definitions so a 4 means the same thing to everyone: 5-point likelihood and impact scales, rating bands, and a 5x5 risk matrix.
Prefer to work in a document? The full register, scoring guide, and worked example are also available in Word format.
Download the .docx versionYou do not start from a blank sheet. Each starter risk comes with a real description, example controls, a suggested owner, and indicative scores. Editing a credible baseline is much faster than composing one from scratch.
| ID | Category | Risk | Inherent score |
|---|---|---|---|
| AI-DL-01 | Data leakage | Customer records pasted into public AI chatbots leave the corporate boundary and may be retained or used for training. | 16 |
| AI-SA-01 | Shadow AI | Staff adopt AI tools IT has never assessed; nobody knows what data flows into them. | 15 |
| AI-AG-01 | Agentic AI | An autonomous agent acts beyond its intended scope: sending messages, changing records, executing transactions. | 15 |
| AI-MB-03 | Model behavior | Model quality degrades through drift or silent vendor updates, and without a baseline nobody notices. | 9 |
Scores shown are indicative inherent scores for a typical mid-size organization with limited AI-specific controls. Re-score every risk against your own environment.
Every risk is scored twice: once before controls (inherent) and once after (residual). The spreadsheet does the math and the color coding; your team supplies the judgment.
Acceptable with routine monitoring. Review at the standard cadence.
Actively mitigate. Assign an owner, agree on a treatment plan, and review at least quarterly.
Escalate to leadership. Treat before further rollout; consider pausing the activity until controls land.
A register is only as good as its inputs. Aona AI discovers every AI tool in use across your organization, shows what data flows into them, and keeps your inventory and risk register audit-ready.