AI Model Wiki

    Skip to main content
    18 min
    Lesson 1 of 4

    AI Risk Taxonomy

    From the course AI Risk Management

    Understanding the full spectrum of AI risks enables comprehensive risk management programs.

    Risk Categories

    Technical Risks

    • Model Performance: Accuracy, reliability, and failure modes
    • Data Quality: Biased, incomplete, or poisoned training data
    • Distribution Shift: Model degradation as real-world data changes
    • Adversarial Attacks: Deliberate attempts to fool AI systems
    • System Integration: Failures at AI-system boundaries

    Operational Risks

    • Dependency Risk: Over-reliance on AI for critical decisions
    • Talent Risk: Shortage of AI expertise for maintenance
    • Vendor Risk: Third-party AI providers changing or discontinuing services
    • Scalability Risk: AI systems failing under load
    • Change Management: Organizational resistance to AI adoption

    Compliance and Legal Risks

    • Regulatory Non-compliance: Violations of AI regulations
    • Intellectual Property: Copyright issues with AI-generated content
    • Liability: Who is responsible for AI-caused harm
    • Privacy Violations: Unauthorized processing of personal data
    • Discrimination: AI systems with disparate impact

    Reputational Risks

    • Public Perception: Negative media coverage of AI failures
    • Bias Incidents: Publicized AI discrimination cases
    • Misinformation: AI-generated false content spread
    • Trust Erosion: Loss of customer confidence in AI systems

    Risk Scoring Framework

    Risk Score = Likelihood × Impact × Velocity
    
    Likelihood Scale (1-5):
    1: Rare (<5% probability)
    2: Unlikely (5-20%)
    3: Possible (20-50%)
    4: Likely (50-80%)
    5: Almost Certain (>80%)
    
    Impact Scale (1-5):
    1: Negligible (no meaningful harm)
    2: Minor (limited impact)
    3: Moderate (significant disruption)
    4: Major (severe harm)
    5: Critical (existential threat)
    
    Velocity Scale (1-3):
    1: Slow (days to weeks to materialize)
    2: Medium (hours to days)
    3: Fast (immediate)
    

    Key Takeaways

    • •AI risks span technical, operational, compliance, and reputational categories
    • •Likelihood × Impact × Velocity scoring enables prioritization of risk management
    • •Distribution shift and adversarial attacks are often underestimated technical risks
    • •Reputational risks from AI failures can have lasting business impact
    0 of 4 lessons marked complete

    Progress is kept in this browser only.