Human oversight still crucial despite AI advances, says expert

 PwC's Victoria Wymark says AI can generate convincing but inaccurate content, making human oversight and stronger governance essential.

Panellists at the International Regulatory Conference 2026 discussed the challenges and future of AI.

KUALA LUMPUR:

Artificial intelligence (AI) can generate convincing content and recommendations, but humans must still review critical decisions because the technology can also produce incorrect answers, says an industry player.

Victoria Wymark, PwC Southeast Asia Consulting’s director of AI and data transformation, said AI does not understand values, ethics or the difference between right and wrong like humans do, but identifies patterns based on data.

“AI can be very creative, but that creativity also means it can generate incorrect answers.

“That’s why important decisions still require scrutiny by humans,” she said at the International Regulatory Conference (IRC) 2026 yesterday.

Wymark also said organisations should not view AI merely as a tool to reduce costs.

“They must also develop robust processes, data management systems and governance frameworks to ensure that the technology is used responsibly,” she said.

CyberSecurity Malaysia head of preemptive technology and services Fazlan Abdullah said advances in AI had made cyber threats far more difficult to detect than they were a few years ago.

He said scam emails, once easily identified by poor grammar and spelling, could now be generated using AI with near-perfect language.

Deepfake videos are also becoming increasingly difficult to distinguish from genuine recordings, he added.

“In the past, we were always told not to believe everything we see. Today, we also cannot immediately trust everything we read.

“AI-generated content is now so convincing that it can deceive users if they are not vigilant,” he said, warning that such threats could also affect businesses and national security.

From a corporate perspective, CelcomDigi head of AI and data innovation Azuree Hydani Misri said the company only allows AI tools that comply with its internal policies on cybersecurity, privacy and data governance.

He said every AI solution involving customer data must undergo security and privacy assessments, as well as testing, before it is approved for use.

Meanwhile, Asean Secretariat senior digital economy officer Hazremi Hamid said AI development in the region must be guided by the principles of transparency, accountability, data governance and a human-centred approach.

He said Asean is strengthening regional cooperation through information sharing, capacity building and AI safety frameworks to help member states manage risks associated with the technology.

Among the initiatives is the Asean AI Safety Network, announced last year as one of the projects under Malaysia’s Asean chairmanship.

Hazremi said the Asean Digital Economy Framework Agreement (Defa) is expected to be signed at the Asean Summit in November to strengthen regional cooperation in the digital economy.

Defa is a proposed regional pact aimed at creating common rules for the digital economy among Asean member states, as well as strengthening cooperation in digital trade, cross-border data flows, online safety and AI governance. - FMT