How to run a risk assessment of an AI system
Run the risk assessment of an AI system step by step: classify the system under the AI Act, describe it, identify and assess the risks to people and to the organisation, then define controls and monitoring. A running case (a recruitment SME deploying a CV-screening tool), spreadsheet and document templates, an option with an AI assistant. Course taught entirely in English.
- 6 chapters
- 28 lessons
- 1 exercise
- 6 quizzes
Graded quizzes, exercises and interactive tools are available in the online course.
Read the first lesson → Create an account Log in to the course
At a glance
- Level
- Intermediate
- Estimated time
- about 9 h
- Audience
- Project managers, lawyers, compliance and risk officers deploying an AI system.
- Recommended first
- Risks and AI
- Learning outcomes
- Classify an AI system under the AI Act.
- Describe the system and identify its risks for people and the organisation.
- Assess the risks and define measures and monitoring.
- Assessment
- A graded quiz for each chapter; in a group, the next chapter opens once the lessons are read and the quiz passed (60% by default, threshold set by the trainer).
- Badge
- Digital badge (Open Badges format) issued by the trainer once every quiz of the course is passed and the final exam passed at 70%: 20 random questions, 30 minutes, at most 3 attempts, 24 h between two attempts.
- Updated
Course outline
Lessons marked “free to read” need no account; the other lessons, the exercises, the quizzes and the answers do.
Chapter 1Framing the analysis4 of 4 lessons free to read
Why an AI-specific analysis, which systems to analyse, and how to qualify them under the AI Act.
- 1.1Why AI systems need their own analysisfree to read
- 1.2Inventory the organisation’s AI systemsfree to read
- 1.3Qualify a system under the AI Actfree to read
- 1.4Roles and obligations: provider, deployer, importerfree to read
Chapter 2Describe the system4 lessons
Purpose, data, components and human in the loop: the description sheet that underpins everything else.
- 2.1Purpose, context of use, affected people
- 2.2The data chain
- 2.3Components: model, provider, hosting, interface
- 2.4Stakeholders and the “human in the loop”
Chapter 3Identify the risks5 lessons
Risk families, fundamental rights, public catalogues and writing scenarios.
- 3.1The risk families of an AI system
- 3.2Risks to fundamental rights
- 3.3Sources: public catalogues of risks and incidents
- 3.4Write risk scenarios
- 3.5Practical sheet: the CV-screening risk list
Chapter 4Evaluate the risks4 lessons
People-centred scales, measures for evaluating, matrix, gross and residual risk, link with the DPIA.
- 4.1Severity and likelihood scales
- 4.2Measure to evaluate: performance, bias, adversarial tests
- 4.3Matrix, prioritisation, gross and residual risk
- 4.4Link with the GDPR impact assessment (DPIA)
Chapter 5Treat and monitor5 lessons
Technical and organisational measures, documentation, post-deployment monitoring, management system.
- 5.1Technical measures
- 5.2Organisational measures
- 5.3Documentation: what the AI Act requires depending on the role
- 5.4Monitoring after entry into service
- 5.5Integrating the analysis into the management system: ISO/IEC 42001 and 23894
Chapter 6Complete step-by-step on the case6 lessons
Redo the whole approach on the Aubépine case, with the templates, first by hand then with an AI.
- 6.1Step 1: qualify the system
- 6.2Step 2: describe the system
- 6.3Step 3: identify the risks (by hand, then with an AI)
- 6.4Step 4: evaluate
- 6.5Step 5: the treatment plan
- 6.6Step 6: the summary sheet