Universities across the world – and in South Africa – are having an id disaster, and generative AI is the set off.
Greater than three years after ChatGPT walked into the lecture corridor, the talk has moved effectively previous dishonest. In a current piece in New York journal, a Columbia undergraduate boasted of utilizing AI to jot down most of his essays; in The New Yorker, the critic Hua Hsu requested, “What occurs after AI destroys faculty writing?” What does a level certify when a chatbot can produce a reliable essay in seconds?
South Africa’s universities will not be watching from the touchline – removed from it. Detailed responses to TechCentral from North-West College (NWU), the College of Pretoria (UP), Stellenbosch College (SU) and the College of the Witwatersrand (Wits) reveal a sector that has deserted an earlier, defensive posture of “AI policing” for one thing extra critical and impressive: a holistic, even philosophical, rethink of educating, evaluation and function.
The clearest signal of the shift is what universities have stopped doing. Stellenbosch “discontinued the usage of Turnitin’s AI textual content detection performance on the finish of 2025”, says Hanelie Adendorff, senior advisor on the college’s Centre for Educating and Studying.
The instruments, she says, don’t work effectively sufficient to justify the hurt they’ll trigger. Impartial evaluations discover their accuracy collapses the second a scholar calmly edits AI-generated textual content, and a extensively cited Stanford examine discovered that essays by non-native English audio system have been wrongly flagged as AI-written 61.3% of the time, towards simply 5.1% for native audio system.
‘From policing to stewardship’
In a rustic the place most college students write, suppose and are assessed in a second or third language, that bias is an issue. Mario Landman of the Tutorial Centre of Excellence, which serves personal training group AdvTech and its Impartial Institute of Training, argues that the sector ought to abandon detection, describing the shift as a transfer “from policing to stewardship”.
Wits’s newly authorized official AI coverage, adopted final month, reaches the identical conclusion from a distinct course: reasonably than police, it asks workers to set “assessments that mirror real-world skilled duties and are inherently troublesome for AI to finish generically”.
Beneath the institutional variations, an necessary consensus is rising. All 4 universities have changed the binary query “Is AI allowed?” with a extra demanding one about what college students ought to have the ability to do with it, and with out it.
Learn: Why AI will get smarter because it scales – a Wits examine has a clue
UP’s tenet is unambiguous: “College students should suppose first, and use AI later,” says the college’s media liaison, Liesel Swart. Stellenbosch is refining an “AI use bar” that helps lecturers determine when AI use needs to be “prohibited, restricted, allowed, inspired or required”.
NWU lets lecturers set the extent of permitted AI per evaluation, “starting from no AI, to brainstorming or enhancing help, to partial or full use”, and the Wits coverage likewise activates “clearly defining the permissible and impermissible makes use of of AI inside particular tutorial duties”.
The second level of settlement is that evaluation itself should change. “The intention isn’t merely to make dishonest tougher, however to make studying extra seen and significant,” says Adendorff, describing a shift at Stellenbosch “from policing AI use retrospectively to fascinated with how we design studying alternatives”.
UP attracts a sharper line between low-stakes coursework meant for follow and suggestions – formative evaluation, within the jargon – and the safe, invigilated exams that really certify a level, which should “confirm that the coed has achieved the required outcomes”.
The Wits coverage requires “AI-resilient evaluation” designed to prioritise originality, important pondering and scholar engagement. Wits can be writing AI into what it assesses. “With the elevated use of AI in trade and work environments, we do want to incorporate the power to critically and ethically use AI in our studying outcomes,” says Nicole de Moist-Billings, its senior director of educational affairs.
Everybody’s enterprise
Third, all 4 deal with AI literacy as everybody’s enterprise, not a distinct segment talent for laptop scientists:
- NWU has run a free course accomplished by greater than 3 000 college students;
- UP embeds AI literacy in a first-year module, knowledgeable by its personal 2025 survey, which discovered that AI use is already mainstream amongst first-years however that accountable use lags behind consciousness of the dangers;
- SU separates “technical” from “skilled” AI literacy and desires the latter in each diploma; and
- The primary of Wits’s six ideas is to “foster AI literacy” for all workers and college students, explicitly “acknowledging that AI entry isn’t equitable in our context”.
There may be additionally a shared insistence that transparency, not surveillance, is the best way ahead: UP expects college students to declare AI use, and Wits attaches an in depth disclosure template to each submission, on the precept that “AI can’t be an writer on any work”.
Not one of the 4 sees AI as an affordable shortcut, both. Requested whether or not it lets them educate extra college students at decrease price, SU and UP answered, in impact, that it’s each: a solution to help college students at scale, and a big new price in licensing, workers improvement and evaluation redesign. At Wits, De Moist-Billings factors to the price of licences to pilot and undertake AI programmes – “new prices that we’ve got not needed to provision for previously” – whereas NWU is probably the most optimistic, anticipating AI tutor bots to chop prices and enhance scholar success.
The colleges diverge most on the place and the way AI needs to be ruled:
- NWU has moved quickest and most formally, changing into what it says is the first college in South Africa and on the continent to approve an official AI coverage, which intentionally locates AI governance contained in the IT division and is run by a devoted AI Hub.
- “Notably, lots of the issues raised have been philosophical reasonably than technical,” Anné Verhoef, the thinker who directs it, mentioned in a press release shared with TechCentral.
- Wits has taken virtually the other route, putting strategic management with its senate and subcommittees, whereas its schools “are main in designing the parameters of use primarily based on particular duties”, says De Moist-Billings. It describes the doc, which handed by way of the college’s senate in November 2025 and was authorized by its council in June 2026, as a “residing” one for a research-intensive establishment.
- UP has anchored its response in senate-approved tips for lecturers reasonably than a single overarching coverage, whereas SU remains to be consolidating “a number of mandatory responses in several environments” right into a extra coherent complete.
They differ, too, on how far to lean on the instruments they’re integrating. SU has switched Turnitin’s AI detector off completely; Wits, UP and NWU lean as a substitute on disclosure and redesign, with NWU including a teaching-first twist: college students suspected of misconduct will not be instantly referred for disciplinary motion however logged on an inner integrity system, “not on their everlasting tutorial file”, given a warning and required to finish a remedial course on accountable AI use.
Learn: AI is breaking the hyperlink between college levels and employment
Even AI literacy is delivered otherwise: NWU retains its flagship scholar course voluntary, “because the intention is to advertise significant engagement reasonably than compliance”; UP builds it right into a required first-year module; and Wits folds it right into a broader first-year digital-literacy course, in order that, in De Moist-Billings’s phrases, “AI literacy isn’t taught in isolation from different literacies”.
Wits sees the hazard the opposite means spherical. The place the others name their greatest threat the hollowing out of a level, De Moist-Billings names the other: “A significant threat for us could be to not interact with AI.”
What’s a college, anyway?
What additional unites the responses is a willingness to let AI reopen the oldest query in increased training. “AI doesn’t take away the necessity for universities. It sharpens the query of what universities are for,” says Swart, insisting a level “should nonetheless certify what the coed can perceive and do, not merely what a software can produce on their behalf”.
Adendorff places it virtually identically: the college “can’t be lowered to producing evaluation outputs or credentials” however exists “to type individuals who can suppose, decide, inquire, create, act ethically and contribute responsibly to society”.
Wits states the stakes on the duvet of its coverage: “Synthetic intelligence won’t outline us. Integrity will.” Verhoef goes additional nonetheless, warning that skilled our bodies could someday open their examinations to self-taught candidates, forcing universities “to rethink their distinctive contribution”.
That soul-searching is unfolding towards a nationwide AI coverage mess. South Africa’s draft nationwide AI coverage was withdrawn not too long ago after it emerged that components of it had been drafted with AI, full with fabricated tutorial citations – a reminder that the temptation to let the machine do the pondering reaches effectively past the coed physique. The job of rebuilding it has since fallen to an knowledgeable panel chaired by Wits AI researcher Benjamin Rosman, head of the college’s Machine Intelligence and Neural Discovery (Thoughts) Institute.
The colleges’ wager is that the reply isn’t higher surveillance however higher design. As Landman places it, the aim “shouldn’t be to win an unwinnable technological race however to ascertain a renewed contract of belief: one through which AI is used as a scaffold for thought, not an alternative choice to it”.
Learn: Malatsi strikes to rescue South Africa’s botched AI coverage
And if the sector will get there, it intends to get there collectively. Requested what units Wits aside from its friends on AI, De Moist-Billings presents a disarming reply: “We’re not doing something completely different from different universities. We’re studying from and with one another.”
The wager, shared throughout all 4 establishments, is {that a} machine that may write the essay nonetheless can’t do all the pondering – and {that a} college’s job, now greater than ever, is to verify its graduates nonetheless can. – © 2026 NewsCentral Media
