L4-M1 / Provider practice
Risk management and data governance
Duplicate CVs cross the train-test boundary and one group is underrepresented. Decide what the headline score can actually establish.
Your fictional workplace case
Strong average accuracy. An uneven reality.
The aggregate result looks promising.
False rejections and data leakage change that picture.
What you will create
A draft: AI risk register and dataset review
Use the fictional case to record evidence, explain a provisional decision and identify the next owner or action. This is a learning draft for self-review, not approved documentation for a real organisation.
See what goes into your worksheet
- Intended purpose, affected people and foreseeable misuse
- Hazard, harm pathway and initial risk rationale
- Dataset source, rights, splits and representativeness
- Mitigation, owner and measurable acceptance criterion
- Test evidence, subgroup results and residual risk
- Release authority, monitoring trigger and review date
From understanding
to better judgement.
Two lessons, a workplace decision and a final check.
Connect hazards to data and design choices
A provider develops a recruitment-ranking tool.
Lesson 1 · 10–15 min02 / LEARNTest subgroups and justify residual risk
The development team reviews apparently strong test results.
Lesson 2 · 10–15 min03 / APPLYApply what you learned
AI risk register and dataset review
Workplace task · 8–12 min04 / CHECKCheck your understanding
Apply the ideas to 15 workplace questions.
Final check · 5–10 minSkills you will practise on the case
A method for your
next work decision.
- Draft an initial risk record tied to intended use and foreseeable misuse.
- Document dataset provenance, suitability and limitations.
- Design evaluations that expose meaningful failures and disparities.
- Connect mitigations, residual risk and release decisions to evidence.
The case is fictional. Timing is a planning estimate; pause whenever you need. This module is a learning activity, not a verified professional certification.