Half of surveyed South Africans said they had bypassed their employers’ restrictions on artificial intelligence tools during the previous year, according to findings presented at a cybersecurity roundtable in Johannesburg.

Prejlin Naidoo, a partner at Oliver Wyman, shared the findings from the Oliver Wyman Forum Global Consumer Survey 2026 South Africa report at the Marsh-hosted event. The report drew on responses from 903 South Africans and is due to be released in November.

Naidoo said 50% of respondents had circumvented enterprise controls on AI tools. She cautioned that the figure was probably underreported, and said managers were more likely than their teams to have engaged in the practice.

“This creates an entirely new kind of threat perimeter,” Naidoo said.

Data and workplace integrity risks

The discussion highlighted how employees can transfer company information into publicly available AI tools, including large language models. Such actions can allow sensitive material to leave an organisation’s controlled environment without the type of conventional cyberattack that companies traditionally monitor.

Naidoo warned that simply supplying approved AI products or blocking unauthorised services may not be enough to manage workplace use. The concern, she said, extends beyond the protection of company data to questions about the reliability and origin of work produced with AI systems.

According to the findings she presented, 35% of people surveyed had seen AI-generated work presented as if it had been created by a person. Naidoo identified the sharing of sensitive company information and the integrity of AI-produced work as two areas requiring close attention.

The survey also found differences between industries. Some 31% of employees in communications, media and technology had used unapproved AI tools, compared with 34% of employees in other industries.

Separate figures from a 2025 Kaspersky study cited at the event indicated that 72.5% of professionals surveyed in South Africa used AI tools for work. Only 30% said they had received training on cybersecurity risks linked to AI. The source material does not provide the Kaspersky study’s sample size or detailed methodology.

Broader cyber resilience concerns

Marsh Africa and South Africa chief executive officer Spiros Fatouros said cyber resilience could not be treated solely as an internal information technology concern. He pointed to incidents involving South African insurance companies in which attackers were able to gain access through suppliers and partners.

“The key message is that cyber resilience can no longer be viewed solely as an internal IT issue,” Fatouros said, arguing that companies also need to consider the security of cloud providers, suppliers, partners and other connected organisations.

Keletjo Chiloane, a career and workforce consulting executive at Marsh, said the spread of AI was also changing the design of jobs. Organisations, she said, would need to review roles, responsibilities and the skills required for people to work with these systems.

“AI is not just a technology implementation. It changes how work gets done, which means organisations need to rethink roles, responsibilities and the skills people need to work alongside these systems,” Chiloane said.

Naidoo also reported that consumer trust in AI had risen in Oliver Wyman’s research from 11% in 2023 to 44% in the latest period referenced. In addition, 41% of those surveyed said they would be comfortable allowing an AI agent to place an order on their behalf.