How to prevent employees from sabotaging AI rollouts

AI may have become ubiquitous in the workplace, but new research shows that some employees are increasingly pushing back on AI tools by sabotaging their rollout.

Generative AI company Writer and research firm Workplace Intelligence found that 29% of 1,200 employees across the UK, US, and Europe had engaged in sabotage. Examples include opting out of AI training, ignoring company guidelines on AI use, and refusing to use tools altogether. There is a greater resistance to AI tools among younger workers, with 44% of Gen Zers admitting to sabotage.

In extreme cases, workers have even resorted to uploading sensitive company data into unapproved tools and tampering with performance metrics to make it seem as though tools are underperforming.

Charles Radclyffe, founder of SaaS enterprise provider EA Global AI, says that workers pushing back in this manner shouldn’t come as a shock. “In many cases, what is being labeled as ‘sabotage’ is simply a rational human response to how AI is being introduced.”

Latest Videos From

Watch full video here:

He adds: “We’ve spent decades asking people to behave like robots – follow processes, fill forms in, and move data from one system to another. Now that the robots can finally do that work, we’re surprised that people push back.”

This view is echoed by the report, which shows that the main reason for the sabotage is that employees don’t want to be replaced by AI (30%). Tools diminishing employees’ value and creativity (26%) and adding to workloads (20%) were other cited reasons.

So, what can leaders do to prevent acts of sabotage from happening?

Be clear on what AI tools will and won’t do

For Radclyffe, the solution is fairly simple: “If your teams are resisting using the tools you adopt, or are using something else, it’s telling you something, so I’d suggest listening. Either the technology isn’t fit for purpose, or the incentives are broken. If you fix those, adoption tends to resolve itself.”

Leaders also need to communicate to their employees that AI is more than a “tooling upgrade”, he adds. If there are implications for roles, then they ought to know.

Adam Gaca, managing director for UK&I at tech strategy adviser and tech partner Future Processing, echoes the importance of communication. “People are being encouraged to use AI more often, but they don’t always have a clear view of where it actually improves outcomes or how their work will be evaluated. In this situation, switching tools, ignoring recommended solutions, or working around the official setup is a fairly natural response,” he says.

“Leaders need to define where AI will be applied, how results are evaluated, and how teams are expected to work with it on a day-to-day basis.”

This will give them reassurance and confidence and means they’d be less likely to push back against rollouts. “Once that’s in place, adoption becomes more consistent and behavior settles naturally,” he adds.

Empower through AI governance

The biggest risk of employees using unauthorized AI tools in the workplace is undoubtedly shadow AI.

Data from the Writer and Workplace Intelligence survey shows 76% of 1,2000 C-suite executives are aware of the dangers posed by AI sabotage and how it could threaten their company’s future. When employees choose to bypass internal AI guardrails and reject approved tools in favor of unauthorized ones, they increase the risk of shadow IT. Nefarious actors can use the unauthorized tools as a backdoor to exploit unpatched systems, leak data, and expand the attack surface.

“Shadow AI often occurs because the approved tools aren’t very good. People will always gravitate towards whatever helps them get the job done faster and better,” says Radcylffe.

The onus is on leaders to establish governance frameworks outlining what constitutes improper use of AI tools and how sensitive data should be handled. Oliver Simonnet, lead cybersecurity researcher at cybersecurity and AI usage control platform CultureAI, says that employees will often avoid approved tools and seek alternatives because “governance has become a blocker rather than an enabler”. In other words, employees are wary of the unknown and need to be empowered to use the approved AI tools.

Leaders should start by positioning AI training for governance compliance – this has been a legal requirement in the EU since 2025, but isn’t yet in the UK – as a career development and learning opportunity, as opposed to a box-ticking exercise. They should also recognize employees for responsible use of approved tools and incentivize them to flag possible bias, hallucinations or data leaks. For example, cash bonuses and rewards could be tied to compliance behaviors.

Ultimately, leaders need to show to employees that the AI tools are being implemented for their benefit and not to their disadvantage. “If employees see AI governance as something that helps them do their jobs better, adoption accelerates. If they see it as something that slows them down, they'll find a way around it,” Simonnet adds.

Place trust in employees

While putting guardrails in place is crucial to prevent the use of unauthorized tools in the workplace, companies shouldn’t necessarily put a blanket ban on tools that sit outside of company policy.

If the above has been followed, then employees shouldn’t fear AI tools being adopted. As a result, it’s less likely they’ll sabotage rollouts, and they should be more inclined to use the tools responsibly and take care to protect sensitive data. They should be trusted to use tools outside of company policy outside of the workplace if they help them to produce better work.

As Ash Gawthorp, CTO at tech consultancy Ten10, puts it: “Trust plays a huge role in whether people engage. People are more likely to engage when they can see how changes support better ways of working and understand the role they continue to play.”