The Rise of Generative AI Challenges in South African Higher Education
Generative Artificial Intelligence (GenAI), tools like ChatGPT and similar large language models (LLMs), have rapidly transformed how students approach learning and assessments across South African universities. Since their widespread adoption around 2023, these technologies have enabled students to generate text, code, and even complex analyses with minimal effort. However, this convenience has sparked significant concerns over academic integrity, prompting institutions to update policies and practices. In early 2026, reports indicate a notable increase in AI-related misconduct, particularly in unproctored assignments, online quizzes, and take-home tasks. Unlike traditional plagiarism, which relies on copied content, GenAI produces seemingly original outputs, making detection more challenging and necessitating a shift from punitive measures to proactive strategies.
South African universities, including leading institutions like the University of Cape Town (UCT), University of the Witwatersrand (Wits), Stellenbosch University (SU), University of Pretoria (UP), University of Johannesburg (UJ), and University of South Africa (Unisa), are at the forefront of this evolution. These establishments enroll over 980,000 students collectively, with higher education participation rates surging by nearly 60% in recent years. The cultural and linguistic diversity of South Africa—11 official languages and varying digital access levels—adds layers of complexity, as AI tools often favor English-dominant datasets, potentially exacerbating inequities.
Documented Surge in AI Misuse Cases
The University of Pretoria (UP) provides concrete data, reporting approximately 53 AI-related disciplinary cases between 2024 and 2025 across exams, tests, assignments, projects, and coding tasks. UJ has noted an uptick in suspected and confirmed cases, mainly in take-home written work and online quizzes, while Unisa has identified incidents through its academic integrity committees, leading to referrals to the Senate. UCT has encountered inappropriate uses, such as undeclared AI-generated submissions, though invigilated exams remain largely unaffected due to strict proctoring.
Wits reports no formal incidents, attributing this to its emphasis on ethical promotion over detection. These cases highlight a pattern: misuse peaks in non-invigilated settings where students can access tools undetected. Professor Sehaam Khan, UJ's Deputy Vice-Chancellor, observes, “With access to AI tools becoming almost ubiquitous... so has the reliance on these technologies for learning and assessment.” This trend mirrors global patterns but is amplified in South Africa by high youth unemployment (over 40%) and pressure to excel academically for job market entry.
UCT's Pioneering AI in Education Framework
Launched in June 2025, UCT's AI in Education Framework marks a progressive stance, prioritizing ethical integration over prohibition. It outlines five core principles: ethical use, critical AI literacies, human-centered education, equity, and adaptive innovation. The university discontinued unreliable AI detectors like Turnitin's AI Score from October 1, 2025, due to false positives and negatives.
Recommendations include mandatory AI literacy training covering tool mechanics, biases, and ethical dilemmas; assessment redesigns like process portfolios and oral defenses; and pilots for AI-enhanced tutoring. For instance, lecturers are encouraged to specify 'traffic-light' rules: green for permitted AI (e.g., grammar checks), amber for conditional use, red for prohibited. This approach fosters 'productive struggle' essential for deep learning while preparing graduates for AI-pervasive workplaces. Read the full framework here.
Wits University's Ethical AI Principles
Wits' AI Framework, aligned with its 2025-2029 Learning and Teaching Strategic Plan, rests on six principles: integrity with disclosure, AI literacy, adaptive practices, human oversight, risk management, and equity. AI use must be acknowledged transparently, and tools cannot claim authorship. Faculty-specific guidelines, such as those from Health Sciences and Science, advise against over-reliance in tests to preserve skill-building.
The focus is augmentative: AI for brainstorming or editing, but humans retain accountability. Training emphasizes biases in multilingual contexts, crucial for Wits' diverse student body. This proactive model has prevented formal misconduct cases, positioning Wits as a leader in responsible innovation. Explore Wits' policies for detailed faculty resources.
Stellenbosch and Pretoria's Assessment Reforms
SU's January 2025 Ethical Use Policy mandates declaration of GenAI via student forms, distinguishing assistive (e.g., Grammarly) from generative tools. It promotes AI literacy courses and resilient assessments like evaluative judgment tasks. UP employs 'triangulated evidence'—drafts, version histories, oral clarifications—eschewing sole reliance on detectors.
- Redesign assessments: Shift to in-person vivas, collaborative projects, and reflective logs.
- Enhance proctoring: AI-enabled tools for online exams.
- Student education: Workshops on boundaries and consequences.
These reforms address SU's multilingual challenges, ensuring AI doesn't widen gaps. UP's 53 cases underscore the urgency, with penalties from warnings to suspensions.
Unisa and UJ's Response to Open Distance Learning Challenges
As Africa's largest distance university, Unisa faces amplified risks in remote assessments. It implements tiered sanctions—education for minor lapses, severe for deception—and invests in proctoring. UJ integrates AI into misconduct codes, stressing fraudulent authorship. Both emphasize transparency, with Unisa's guidelines aligning with its Academic Integrity Policy.
| University | Key Strategy | Cases Reported |
|---|---|---|
| UP | Triangulation & reform | 53 (2024-25) |
| Unisa | Tiered consequences | Increasing |
| UJ | Disclosure rules | Uptick |
National Coordination via Universities South Africa
Universities South Africa (USAf) compiled institutional policies in June 2025, highlighting themes like stakeholder consultation, training, and assessment scales (e.g., Perkins AI Scale). Institutions like NWU and UWC embed AI in integrity policies, promoting agile, context-specific guidelines. This collaboration ensures balanced, multi-perspective approaches amid resource disparities. View the USAf compilation.
Challenges: Equity, Detection, and Cultural Context
Digital divides persist: Rural students lack reliable internet, hindering ethical AI access. Biases in English-centric models disadvantage isiZulu or Afrikaans speakers. Detection tools' unreliability (e.g., 20-30% error rates) risks unfair accusations. Experts like Prof. Boitumelo Senokoane (Unisa) advocate education: “Many incidents involve students who did not fully appreciate the boundaries.” Solutions include subsidized tools and multilingual training.
Expert Perspectives and Stakeholder Views
Prof. Jonathan Jansen emphasizes oral assessments post-assignments. Librarians and IT specialists contribute to policy drafting. Students, via co-creation at UCT/SU, favor guidelines over bans. Faculty worry about workload but value innovation pilots. This multi-stakeholder dialogue ensures policies evolve dynamically.
Photo by Jolame Chirwa on Unsplash
Future Outlook: Towards AI-Resilient Higher Education
By 2030, SA aims for AI-integrated curricula per National Development Plan. Universities plan scaled training, enterprise tools (e.g., licensed Gemini), and research on local biases. Implications: Enhanced graduate employability in 4IR jobs, preserved integrity signaling quality degrees. Actionable insights for educators: Start with literacy modules; for students, declare usage; for admins, fund redesigns.
These updates position South African universities as adaptive leaders, balancing integrity with technological progress for a knowledgeable society.