Stellenbosch University in South Africa has rejected a single, campus-wide rule for artificial intelligence, choosing instead to let individual professors set their own guidelines. Under this new policy, educators can allow, restrict, or require generative AI depending on the goals of each assignment. This flexible model is a change from how higher education institutions usually handle emerging technology.
What Happened
On January 1, 2025, Stellenbosch University implemented a new Senate-approved ethical position statement on AI, according to EdTech Innovation Hub. Instead of a blanket ban, the university's new AI in Teaching, Learning and Assessment resource gives educators the authority to decide if tools like ChatGPT fit their course goals. Lecturers must state their AI rules and explain the educational reasons behind them.
Importantly, students remain fully responsible for the accuracy and originality of their work. If an educator suspects a student used AI inappropriately, the university warns against relying solely on AI detection software. Instead, professors are encouraged to talk with students or conduct oral assessments to verify that learning actually took place.
The Bigger Picture
Stellenbosch is not alone in shifting away from absolute bans. A global study of 21 English-medium universities published in Participatory Educational Research found that institutions are increasingly moving away from blanket prohibitions and toward "assessment-aware" guidance. Some universities are adapting by using "process-evidence portfolios" and in-class oral verifications to make sure students actually understand their work.
This transition is driven partly by the unreliability of AI detection tools. While some universities focus on policing, researchers at the University of Florida found that commercial AI detectors are poorly suited for high-stakes academic settings. Their study revealed that minor changes to AI-generated text make detectors ineffective, leading to false-negative rates as high as 99.6% and false-positive rates up to 68.6%. A separate study in ScienceDirect confirmed that the technology gap favors AI creators over detector developers, making automated proof of cheating highly flawed. Even when detection tools successfully flag AI text, researchers writing for the Vrije Universiteit Brussel warn they should only be initial flags rather than definitive proof of cheating.
What This Means for Families
For parents and students, this decentralized shift means rules will change from classroom to classroom. A student might be required to use AI for a computer science project, but completely forbidden from using it for a history essay. This mirrors global trends in educational technology, where schools must find a flexible balance, similar to how regional hubs must tailor local EdTech systems to regional needs.
Students must learn the difference between basic technical help (like using spelling tools) and substantive contributions (like letting AI write entire paragraphs). Academic standards are increasingly demanding that students explicitly disclose AI assistance rather than just citing it like a traditional book.
What You Can Do
- Students should check the syllabus for every class. Since rules are decentralized, students must review the specific AI guidelines for each course rather than assuming a single university policy.
- Students should keep a prompt log. When using AI for brainstorming or drafting, saving prompts and drafts protects against false cheating accusations by showing how the work evolved, as recommended by LogAI.
- Students should prepare a clear AI declaration when using these tools. They should write a statement detailing which tool was used, how it helped, and how they verified the accuracy of the final output, a practice supported by Cornell University guidelines.