WHAT CAN EACH AI
SYSTEM GET WRONG?
A structured look at the AI systems you build or buy: what data they touch, how they can fail and what that would cost you.
| System | Risk | Likelihood | Impact |
|---|---|---|---|
| Customer-support chatbot | Data leakage | Medium | High |
| CV screening tool | Bias | Medium | High |
| Document summariser | Hallucination | High | Medium |
| Internal search assistant | Access control | Low | Medium |
Frameworks and standards
- ISO 42001
- Australian AI Safety Standard
- Privacy Act
Typically for
- Head of risk
- CISO
- Product and data leaders
THE DELIVERABLES.
AI risk register
Data leakage, bias, hallucination impact, supplier exposure and IP risk, recorded per system.
Consistent ratings
Likelihood and impact scored the same way every time, so you can compare systems and decide where to spend.
Treatment plan
Prioritised actions with owners, from quick fixes to design changes.
Executive summary
A one-page view for your board or risk committee.
THREE STEPS.
Here's what happens, in order.
Scope
Pick the systems that matter: third-party tools, in-house models and AI embedded in software you already own.
Assess
Walk through data flows, failure modes, controls and supplier terms with the people who run each system.
Treat
Agree the plan and hand over a register your team can keep up to date.
A GOOD FIT IF…
- You deploy AI that makes or influences decisions about people.
- You use third-party AI on customer or employee data.
- Your board, insurer or customers want an AI risk view.
QUICK ANSWERS.
What is an AI risk assessment?
It evaluates the risks of your AI systems, whether third-party tools like ChatGPT and Copilot or your own models: data leakage, bias, hallucination, supplier weaknesses and regulatory gaps. You get a prioritised treatment plan.
Do you test models technically?
We assess risk, controls and suppliers. Where deeper technical testing is needed we scope it and bring in specialist testers, and we tell you who they are.
TALK TO A
SENIOR PERSON.
Tell us where you are. We'll reply within one business day.
Talk about AI Risk Assessments ↗