Lesson 10 of 12
Structured learning draftRisk review with Red-teaming
In Responsible AI Development, the way a learner handles risk review shapes how Red-teaming is used and evaluated. Risk review prioritises harms and tests controls. This intermediate lesson focuses on a decision or output that another person can inspect.
Learning objectives
- Explain risk review in the context of Responsible AI Development.
- Apply Red-teaming to a bounded practical task.
- Evaluate the result using explicit quality criteria.
Risk Review: from context to evidence
Risk Review connects affected people and context to a documented mitigation in Responsible AI Development.
Define the purpose, intended user and Red-teaming constraints.
Use affected-stakeholder and technical evidence.
Compare the observed result with a normal case, boundary case and stated limitation.
Review the evidence, adjust the method, and repeat.
Risk review prioritises harms and tests controls. For Red-teaming, distinguish performing an operation from demonstrating that it suits the stated purpose. Use affected-stakeholder and technical evidence. Record assumptions that could change the conclusion.
Apply risk review deliberately
- State the Responsible AI Development task and the decision it supports.
- Prepare a small Red-teaming case with a known input and difficult boundary.
- Use affected-stakeholder and technical evidence.
- Compare the observed result with the expected behaviour and explain differences.
- Save the evidence, limitation and next action in a review record.
A worked Red-teaming evidence path
A four-step worked example for applying risk review to Red-teaming, including a boundary test and revision.
Preserve the original Red-teaming case and expected result.
Confirm the basic path behaves as expected.
Expose an assumption in the risk review method.
Change the method, rerun both cases and record the limitation.
| Review point | Evidence |
|---|---|
| Purpose | The specific Red-teaming outcome and intended user |
| Method | The risk review decision, input and version or context |
| Result | Observed output plus a checked boundary case |
| Limitation | What the result does not establish and the next safe action |
Common mistakes
- Using Red-teaming before defining what risk review must achieve.
- Checking only the easiest Responsible AI Development example.
- Reporting a result without its input, assumptions or limitation.
Practice activity
Apply the lesson
For Responsible AI Development, complete a bounded Red-teaming task demonstrating risk review. Keep the original input, numbered method, normal test, boundary test, observed results and a 100-word self-review naming one limitation and next improvement.
Check your understanding
In Responsible AI Development, which evidence best supports a risk review result produced with Red-teaming?
Lesson summary
- For Responsible AI Development, risk review means: Risk review prioritises harms and tests controls.
- A credible Red-teaming result includes a checked boundary, not only a successful example.
- The next lesson builds on this risk review evidence record.
Sources and further reading
- Recommendation on the Ethics of AIUNESCO - accessed 2026-08-21
- AI PrinciplesOECD - accessed 2026-08-21
Personal study note