Module 3: Governance & Risk Management, BoK III.A

Calculating risk

A simple likelihood-impact matrix can multiply probability and severity scores. Select a response using the applicable law, risk tolerance, available mitigation and potential impacts. A high score alone does not universally require abandonment.

The process

Identify the affected people, business purpose, planned use and plausible harms. Assess likelihood and severity using a documented method. Then decide whether to avoid, modify, mitigate or accept the risk within legal requirements and the organization's risk tolerance. An unacceptable residual risk can require stopping or changing the use.

  • Intended task → what will it achieve
  • Function & performance metrics
  • Robustness → scalable, withstands more or less use
  • Transparency → workings and intended consequences
Fairness is at the core

All assessment criteria sum to one question → is the AI outcome ultimately fair?

The EU risk pyramid
TierWhat it means
Unacceptable riskSpecified practices prohibited under EU AI Act Article 5, subject to the detailed scope, conditions and exceptions.
High riskHarms safety or fundamental rights → mandatory requirements apply
Limited riskSpecified transparency duties may apply. Check the particular provision and other applicable law.
Minimal riskThe simplified minimal-risk label does not remove duties under other applicable law. Some AI Act provisions may also apply independently of high-risk classification.

Risk assessment is context-specific → context comes from five places → owner and operator, industry and use case, social impacts, timing, jurisdiction. Tailoring the framework to context strengthens governance and enables informed, values-aligned decisions.

Practitioner voice

"Risk assessment of AI is both an art and a science." - Vivienne Artz

Exam flash

Choose a response from the actual facts. Unacceptable risk without adequate mitigation can require avoiding or changing the use. The EU AI Act separately prohibits specified practices under Article 5. Not all social scoring or all facial recognition is universally prohibited.

Key terms - quick answers

What is “Risk formula”?
Probability of the harm × potential (severity) of it happening.
What is “EU risk pyramid”?
Four tiers, unacceptable (banned), high (mandatory requirements), limited (transparency), minimal (voluntary).
What is “Unacceptable risk”?
Banned AI uses such as social credit scoring and real-time remote facial recognition in public spaces.

Sources and study method

This independent lesson uses active recall, spaced retrieval and scenario practice. Read the full study method.