The Growing Challenge of AI Misuse in South African Higher Education

South African universities are facing an unprecedented wave of artificial intelligence (AI) misuse by students in academic assignments, prompting a coordinated clampdown across institutions. Generative AI tools, such as ChatGPT, Google Gemini, and Microsoft Copilot, have become easily accessible via smartphones and browsers, enabling students to generate essays, code, and answers with minimal effort. This shift has blurred the lines between legitimate assistance and outright cheating, often termed 'aigiarism' or fraudulent authorship. Institutions report a surge in cases particularly in take-home assignments, online quizzes, and coding tasks, while invigilated exams remain less affected due to strict monitoring.

The crisis escalated notably in 2025 and into 2026, with the University of South Africa (UNISA) issuing stark warnings about an integrity crisis fueled by AI-assisted plagiarism. This phenomenon undermines the value of degrees, erodes trust in academic credentials, and overloads disciplinary systems. Yet, amid the challenges, universities are pioneering balanced approaches that integrate AI ethically while safeguarding standards.

UNISA Leads the Charge with Zero-Tolerance Stance

UNISA, Africa's largest distance-learning provider, has been at the forefront of the AI cheating surge. In February 2026, the institution highlighted record levels of AI misuse in exams and assignments, contributing to plagiarism spikes and a burgeoning backlog of disciplinary cases. The university employs a multi-pronged strategy, including AI-enabled proctoring tools like the Invigilator App, Moodle quizzes, and IRIS for behavioral monitoring during online assessments. Turnitin, a web-based similarity checker with an integrated AI content detector, is heavily promoted to flag suspicious linguistic patterns and statistical anomalies indicative of machine-generated text.

UNISA's policy maintains zero tolerance for academic dishonesty, treating unauthorized AI use as misconduct. Consequences range from withheld marks or zero scores for minor first offenses to suspensions, module failures, and even degree revocation for repeat violations. Acting Executive Director Professor Boitumelo Senokoane emphasized, 'The university has a zero tolerance for any forms of academic dishonesty.' To foster ethical habits, UNISA introduced a compulsory Academic Integrity course in 2025 and is finalizing a comprehensive AI policy, encouraging disclosure and responsible use while allowing students to continue studies pending investigations under the 'innocent until proven guilty' principle.

University of Pretoria's Disciplinary Surge and Evidence-Based Approach

The University of Pretoria (UP) documented approximately 53 AI-related disciplinary matters across 2024 and 2025, channeling cases to its legal department for fair adjudication. Common infractions include failing to declare AI use where required or submitting fully AI-generated content misrepresented as original. UP employs a 'traffic-light' system in assignment briefs—green for permitted, yellow for conditional, red for prohibited—clarifying expectations upfront.

Detection relies on triangulation: combining student declarations, draft histories, working notes, oral defenses, and expert reviews rather than sole dependence on tools like Turnitin, which spokesperson Rikus Delport deems unreliable for high-stakes decisions due to false positives and negatives. Sanctions are proportionate, escalating with intent and severity, aligning with broader academic integrity procedures.

Divergent Policies: From UJ's Fraudulent Authorship to UCT's Ethical Shift

The University of Johannesburg (UJ) classifies undisclosed AI use as fraudulent authorship, handling it under existing misconduct frameworks with tiered sanctions based on study level and intent. Deputy Vice-Chancellor Professor Sehaam Khan notes, 'Our policy developments explicitly recognise that the core offence in AI misuse is fraudulent authorship and failure to disclose, not the mere presence of technology.' UJ concentrates efforts on vulnerable formats like take-home work.

In contrast, the University of Cape Town (UCT) discontinued Turnitin's AI Score feature on October 1, 2025, citing its proneness to errors and potential to foster adversarial relations. UCT's June 2025 AI in Education Framework prioritizes human-centered principles: ethical use, critical literacies (e.g., understanding biases, prompt engineering), equity, and innovation. Assessments demand declarations, with misconduct for unauthorized generative AI. Spokesperson Elijah Moholola affirms, 'So-called AI detectors are widely regarded as unreliable.'UCT's framework guides multipronged evaluations like portfolios and orals.

Stellenbosch and NWU Pioneer Guideline Frameworks

Stellenbosch University (SU) issued interim guidelines emphasizing four pillars: accountability (verify outputs), authenticity (defend in own words), transparency (declare via forms/screenshots), and fairness (equitable access). Allowable uses include brainstorming or editing akin to spellcheckers, but prohibitions cover full generation or paraphrasing non-original text. Lecturers must specify rules, model ethical use, and redesign assessments (e.g., orals) to minimize AI utility. Turnitin AI scores serve as cautions only, not proof.SU's guidelines stress learning over punishment.

North-West University (NWU) made history in November 2025 as South Africa's first with an official AI policy, human-centered and sustainable across teaching, research, and governance. It sets national standards for ethical deployment, addressing risks like bias and privacy while harnessing benefits.

Photo by Jolame Chirwa on Unsplash

The Pitfalls of AI Detection Tools

• High false positive rates (e.g., 1% for Turnitin), wrongly accusing non-native English speakers or edited text.
• False negatives miss up to 15% of AI content, especially short or multilingual outputs.
• Easily bypassed via paraphrasing or hybrid human-AI writing.
• Undermine trust; UCT, UP, SU advise against sole reliance.

Institutions now favor holistic verification: version histories, peer reviews, viva voces, and behavioral proctoring. This evolution reflects global critiques, prioritizing fairness over flawed tech.

Redesigning Assessments for the AI Era

To thwart misuse, universities are overhauling evaluations:

Student Perspectives and Ethical AI Literacies

Surveys reveal nuanced views: many master's students use AI for comprehension (41% laptops, 30% smartphones), not evasion. Yet, ambiguity breeds unintentional violations. Unis counter with literacies training—prompt crafting, bias detection, citation—embedding them in curricula. UNISA's integrity module exemplifies proactive education.

Broader Implications for Academic Integrity and Employability

Unchecked AI erodes degree credibility, vital in South Africa's 30%+ youth unemployment context. Stakeholders—from lecturers burdened by redesigns to employers questioning skills—demand action. Balanced policies preserve rigor while preparing graduates for AI-driven workplaces.

Future Outlook: Towards a National Framework

With NWU/UCT models gaining traction, a national AI guideline looms. Unis advocate collaboration: infrastructure investment, faculty training, student buy-in. By 2026, South Africa positions as ethical AI leader in African higher education, blending innovation with integrity.ITWeb analysis forecasts sustained adaptation.

Photo by Jolame Chirwa on Unsplash

Actionable Insights for Students and Educators

• Students: Declare AI, maintain records, practice orals.
• Educators: Specify rules, diversify assessments, teach literacies.
• Institutions: Invest in training, monitor trends.

This clampdown heralds a resilient higher education landscape.