about a third Employees surveyed by a global study have admitted to deliberately undermining their companies' AI roll-outs – and while the research did not include South African respondents, a local AI adoption expert tells TechCentral that resistance is visible here too, though how widespread it is depends largely on company culture.
Study The study, commissioned by enterprise AI vendor Writer and conducted by Workplace Intelligence, of 2,400 knowledge workers in the US, UK and Europe, found that 29% of employees admit to actively sabotaging their company's AI roll-out. Among Gen Z workers, that figure rises to 44%.
Reported subversion takes many forms: injecting proprietary data into public AI tools, using non-approved platforms, intentionally producing low-quality AI output, and, in some cases, tampering with performance metrics to make AI appear ineffective. The most cited motive, according to the 30% of respondents admitting to such behavior, is the fear of losing their job.
It's worth noting that Righter is a commercial AI platform vendor, and while the research it commissioned concluded that executives should move faster on AI adoption, it should be read with that context in mind. What management calls sabotage, employees may describe as caution about devices that still cause hallucinations, misrepresent sources or expose sensitive data if used without adequate controls.
Much of what was described as sabotage in the author's study overlaps with a phenomenon the industry has dubbed “shadow AI” – employees using consumer AI tools like ChatGPT, Cloud, or Gemini outside their organization's accepted technology stack, often because the accepted tools are slower, less capable, or not yet deployed.
Chaya Aye
Shadow AI is a real security and governance concern, especially when proprietary or customer data is pasted into public models with uncertain data retention policies. But this is not direct sabotage.
In many cases, this reflects employees going to work using the best tools available to them, ahead of their employers' procurement and risk processes. Whether this is a problem that needs to be addressed or a sign that official AI roll-outs are lagging behind user demand depends largely on which side of the desk one is sitting.
Reading: South Africa's draft AI policy is a bureaucrat's dream
South African AI expert dean furmanCEO and founder of consulting and training firm 1064 Degrees and author exponential potentialSaid that opposition to AI tools by employees in South Africa is real but not uniform.
“It's not typically widespread. It's not like the default. It's very much dependent on the company culture,” Furman said in an interview with TechCentral.
A qualified actuary and former executive at Discovery and Alex Forbes, Furman has been training South African corporates on AI adoption for several years and has previously appeared as a guest on AI. Techcentral Show.
He argued that culture matters more than territory. “In some companies, there is a lot of psychological safety. They know that this is a company that puts its people first. But then there are other companies where they don't have that safety, and they know that the leaders are a little more ruthless. And there, a lot of the resistance comes, because in that situation, they don't want to show that AI is so useful, because then it makes them seem less valuable.”
Resistance also manifests in different ways across seniority levels. Furman said senior professionals – lawyers and actuaries – often object because of identity. “They almost had a mental breakdown in saying, 'Well, we've got so much training and so much information that AI can never be as good as us at certain things', which is nonsense, because it exceeds their capabilities in many ways.”
Further down the hierarchy, the calculations are more self-interested. “Leaders will be so excited, oh wow, something that previously took two days can now take 20 minutes. But from those individuals' perspective, it's a case of, 'Oh well, I really don't want my managers to know I can do it in 20 minutes now, because then I'm less valuable.'”
The business costs of that thinking are significant, Furman said. “If the process was previously taking a day and now it's possible to do that process in an hour, then the remaining seven hours of the workday is really an opportunity cost for a particular day. So again, if people aren't accepting it, that's the cost.”
The leadership often turns a blind eye to the problem and part of the responsibility falls on the officers themselves, he said. “It's important for leaders to understand AI themselves. Because if not, they won't know that a certain process will now take much less time. There's no way to find out, because they don't know what to benchmark it against.”
underlying trend
The author and the Workplace Intelligence research include a number of findings that will spark debate for employees and executives alike – although these figures also reflect a global survey rather than South African conditions.
According to the report, self-identified AI “super users” are nearly three times more likely to receive both a promotion and a pay raise than non-users. Employees using AI tools reportedly save about six hours per week, while executives save about 12 hours. The report also claims that 60% of executives are planning to fire employees who can't or won't use AI, 69% of companies are already making AI-related layoffs, and 76% of C-suite respondents consider employee sabotage a serious threat to their company's future.
Reading: Rishi claims AI can save small business owners from administrative hell
Those figures are self-reported survey data and should be read as such. In particular, the causal relationship between AI use and promotion is one that the report claims rather than establishes.
Furman said the underlying trend in South Africa is changing, even if resistance remains. “I definitely think people are starting to realize that this is not just a phase. This is a new way of working. AI is not going anywhere.” Areas of deep resistance remain, he added: “There are certain individuals who have a lot of fear.”
For South African officials, the useful question is not whether the global subversion numbers apply here – they may or may not – but rather whether their own organizations have the psychological safety, leadership AI literacy and honest measurements to know the difference between real deterrence and reasonable precautions. – © 2026 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here.