Home / Glossary / AI risk assessment
Glossary
What is an AI risk assessment?
An AI risk assessment is a structured review of the harm an AI system could cause, how likely that harm is and what the company will do about it. It looks at the data the AI touches, the systems it can reach and the actions it can take. The result is a written record with a decision for each risk.
By the AIBound team · Updated 17 November 2026
Key takeaways
- An AI risk assessment reviews what an AI system could harm, how likely that is and how the company will respond.
- It needs facts about the AI in use, including who runs it and what it can reach.
- The written result is evidence that auditors, customers and boards ask to see.
Why an AI risk assessment matters
A company that uses AI takes on new risks. Data can leave through a prompt. An agent can change a system nobody meant it to touch. An AI risk assessment names those risks before they turn into incidents and records what the company decided to do.
It is also what outsiders ask for. Auditors and customers want to see that AI risks were reviewed. The NIST AI Risk Management Framework, a voluntary framework from the US standards body, organises this work into four functions: Govern, Map, Measure and Manage.
How an AI risk assessment works
An AI risk assessment follows the same steps as other risk reviews. The first step is often the hardest, because much AI is installed by employees and never reported.
- 1 Identify the AI system or tool, who uses it and what it is used for.
- 2 Map what it can reach: the data, the accounts and the company systems.
- 3 List what could go wrong, and rate how likely and how serious each risk is.
- 4 Decide the response for each risk, record it and set a date to review it.
AI risk assessment vs an AI audit
An AI risk assessment looks ahead at what could go wrong. An AI audit looks back and checks whether the company did what its rules and its assessments said it would do.
| Question | AI risk assessment | AI audit |
|---|---|---|
| What does it ask? | What could go wrong, and how do we respond | Did the company follow its rules |
| Who runs it? | The risk, security or GRC team | An internal or outside auditor |
| What does it produce? | A list of risks with a decision for each | Findings on what was and was not done |
Sources
- NIST AI Risk Management Framework · nist.gov
- NIST AI RMF Core · airc.nist.gov
Sources checked 17 November 2026
Examples
AI risk assessment examples
FAQ
Frequently asked questions
What is an AI risk assessment?
An AI risk assessment is a structured review of the harm an AI system could cause, how likely that harm is and how the company will respond. It covers the data the AI touches, the systems it can reach and the actions it can take. It ends with a written record.
What are the steps in an AI risk assessment?
There are four steps. Identify the AI and what it is used for. Map the data and systems it can reach. List the risks and rate how likely and how serious each one is. Then decide the response for each risk, write it down and set a review date.
Who should carry out an AI risk assessment?
It is shared work. The risk or GRC team usually leads and keeps the record. Security adds what the AI can reach and how it could be misused. Legal and privacy teams check duties under law. The business owner of the AI explains what it is used for and accepts the result.
How does AIBound support an AI risk assessment?
AIBound supplies the facts an assessment needs. AISeeker finds the AI on employee devices, AgentTracer maps what each agent can reach, and each finding is graded A to F. AIAssure gives each risk a report with the evidence behind it and a clear recommendation, ready to download for audit.