DEMO AI SOLUTIONS AB
AI SYSTEM RISK REGISTER
Document ID: DEMOAI-REG-RISK-001
Version: 3.1 | Updated by: Erik Johansson, Risk Manager | Date: 2026-04-10

=== FICTIONAL DOCUMENT — FOR DEMONSTRATION PURPOSES ONLY ===

RISK ESTIMATION METHODOLOGY

Semi-quantitative risk estimation using likelihood (1-5) and severity (1-5)
scales. Risk evaluation results are categorized as Low (1-6), Medium (7-12),
High (13-19), Critical (20-25).

IDENTIFIED RISKS

RISK-001: False negative — defective product passes inspection
  Category:    Safety
  Description: The classifier fails to detect a manufacturing defect,
               allowing a defective component to pass quality control.
  Method:      FMEA analysis session 2025-11-20
  Likelihood:  3 | Severity: 5 | Score: 15 (HIGH)
  Status:      Mitigated (MITI-001, MITI-002)

RISK-002: False positive — functional product rejected
  Category:    Operational
  Description: The classifier incorrectly flags a non-defective component
               as defective, causing unnecessary rejection and waste.
  Method:      FMEA analysis session 2025-11-20
  Likelihood:  4 | Severity: 2 | Score: 8 (MEDIUM)
  Status:      Mitigated (MITI-003)

RISK-003: Model drift — accuracy degradation over time
  Category:    Performance
  Description: Changes in manufacturing processes or lighting conditions
               cause the model accuracy to degrade below acceptable levels.
  Method:      Literature review and expert consultation
  Likelihood:  4 | Severity: 4 | Score: 16 (HIGH)
  Status:      Mitigated (MITI-004)

RISK-004: Bias in training data — underrepresentation of defect types
  Category:    Fairness
  Description: Training dataset may not represent all possible defect types,
               leading to systematic blind spots in detection.
  Method:      Data quality audit 2025-12-05
  Likelihood:  3 | Severity: 3 | Score: 9 (MEDIUM)
  Status:      Accepted (MITI-005)

RISK-005: Worker over-reliance on automated system
  Category:    Human factors
  Description: Quality control operators may reduce manual inspection
               vigilance due to over-reliance on the AI system.
  Method:      Workplace observation and operator interviews
  Likelihood:  3 | Severity: 4 | Score: 12 (MEDIUM)
  Status:      Mitigated (MITI-006)