Using AI Register
Record risks and incidents
Keep a risk register with scores and treatments, and log what goes wrong so you can learn from it.
The risk register
A risk is something that could go wrong with your use of AI. Open Risk register and click New risk.
- Give it a short title, for example "Staff paste patient details into ChatGPT".
- Link it to an AI system, or leave it blank for a business-wide risk.
- Choose a category: privacy, bias, accuracy, security, transparency, copyright and confidentiality, health and safety, vendor, or other.
- Rate its likelihood and consequence from 1 to 5. The score and its level (High, Medium or Low) appear as you choose. See how risk scores work.
- Record the treatment, what you'll do about it, plus an owner, a status and a due date.
Every high risk should have a treatment plan, an owner and a due date, or a documented decision to accept it.
The list is sorted by score, so your biggest risks are at the top. Filter by status, category, AI system, or high risks only.
The incident log
An incident is something that actually went wrong, or nearly did. Examples are a chatbot giving a wrong answer, a data leak, a complaint, or a harmful output.
Open Incident log and click New incident. Record:
- the date and a short title;
- the AI system involved, if any;
- its severity: low, medium, high or critical;
- what happened, and the actions you took; and
- its status: open, investigating or resolved.
After an incident, check whether a related risk needs a higher score or a new treatment.
Write it up
For a clear, shareable record of an incident, use AI documents → Incident write-up. See Draft policies, assessments and reports.
If personal information may have been involved, check whether you need to notify under the Notifiable Data Breaches scheme. The Office of the Australian Information Commissioner (OAIC) has guidance.
AI Register gives guidance only, not legal advice. Still stuck? Contact us.