Sector Tracks — Module 7A through 7E

South Africa's DCDT published a national AI policy draft with fabricated academic citations generated by AI. The professionals who wrote it were not careless — they lacked the framework to know when to question AI output.

• Assess the risk profile of 8 common corporate AI use cases using a structured framework
• Identify which SA regulations (POPIA, CPA, EEA, B-BBEE) apply to their AI deployments
• Evaluate AI tools (Copilot, ChatGPT, DeepSeek) against corporate risk criteria
• Present a structured AI judgment recommendation to a board or senior team

Millions of SASSA Social Relief of Distress grant applications are declined by an automated system with the message "Alternative Income Source Identified" — often incorrectly. Applicants have no clear path to challenge the decision.

• Identify when automated government decisions require human review under POPIA Section 71
• Map the accountability chain for AI-assisted decisions in their department
• Apply PAIA, PEPUDA, and PFMA considerations to AI procurement and deployment
• Design a basic override and escalation protocol for automated decisions

51% of South African adults are effectively invisible to formal AI credit scoring models. The informal economy — stokvels, cash income, township enterprise — generates no data trail that credit AI can read, creating structural exclusion that mirrors apartheid-era financial barriers.

• Identify proxy discrimination in AI credit and fraud scoring outputs
• Apply NCA Sections 80–81 and FSCA guidance to AI-assisted lending decisions
• Evaluate risk profiles of major SA fintech AI tools (ABSA Abby, JUMO, FNB fraud detection)
• Use the Funda Five to decide whether to act on an AI fraud flag

Mediclinic implemented an AI-assisted workforce management system, followed by a hiring freeze and restructuring. After public backlash, the company denied AI was responsible for staffing decisions — while the AI remained in use.

• Evaluate AI clinical decision support tools using patient safety and equity criteria
• Apply HPCSA Booklet 20 and SAHPRA AI/ML guidance to tool assessment
• Identify POPIA obligations for health data processed by AI systems
• Use the Funda Five to decide when to override a clinical AI recommendation

UCT formally discontinued Turnitin's AI detection tool after evidence of false positives at significantly higher rates for non-native English speakers — disproportionately affecting African students. The decision followed a wave of contested academic integrity rulings.

• Evaluate AI detection and learning tools for equity and false positive risk
• Apply POPIA obligations for learner data and HEQC/Umalusi accreditation considerations
• Identify the accountability chain when an AI detection tool flags a student
• Use the Funda Five to assess an edtech platform pitch for their institution