Flexibility needed as AI outpaces regulations, say experts

As artificial intelligence systems evolve rapidly, traditional regulatory policies designed for an analogue world often struggle to keep pace, leaving both users and innovators navigating through uncertainty.

Experts say rules to govern the use of AI must evolve to adapt to and meet prevailing risks. (Rawpixel pic)

PETALING JAYA:

 Experts have called on the government to move beyond blanket bans in regulating artificial intelligence (AI), recommending a tiered, risk-based framework that protects the public while encouraging innovation.

They said blanket bans on AI tools — such as the recent restriction on artificial intelligence chatbot Grok AI — should be treated only as stopgap measures rather than long-term solutions.

Given the rapid pace of AI development, compared with policies that often take years to formulate, they said regulations must be flexible and adaptive.

As AI systems evolve rapidly, traditional regulatory policies designed for an analogue world often struggle to keep pace, leaving both users and innovators navigating through uncertainty.

Cybersecurity expert Selvakumar Manickam of Universiti Sains Malaysia said AI systems should be categorised according to the level of risk they pose.

“Malaysia should adopt a tiered approach to AI regulation, where systems are categorised based on their potential impact and risk,” Selvakumar told FMT.

He said high-risk systems, such as those capable of generating exploitative content, should be subjected to rigorous pre-deployment testing, continuous monitoring, and independent audits.

Medium-risk applications, meanwhile, should be subjected to transparency obligations, while low-risk tools could operate under voluntary codes of conduct.

“Instead of a static approval process, AI providers should be governed by prevailing obligations, including mandatory incident reporting whenever a system causes or risks causing harm,” he said.

This, he added, would ensure oversight remains dynamic, with regular audits of AI behaviour to address emerging risks in real time, rather than relying on initial assessments that may quickly become outdated as technology evolves.

Recently, Malaysia imposed a temporary restriction on artificial intelligence chatbot Grok in response to concerns over the chatbot’s ability to generate offensive and harmful content.

Last week, the Malaysian Communications and Multimedia Commission (MCMC) announced that the ban on Grok had been lifted after social media platform X confirmed that its artificial intelligence chatbot could not be misused to generate or edit harmful content such as pornography, sexual material or incest.

Farlina Said of the Institute of Strategic and International Studies Malaysia said extended bans that fail to address core issues can be unhealthy for the AI ecosystem in the long run.

According to Farlina, over-reliance on prohibitions risks stifling innovation and limits opportunities for regulators to learn and improve oversight mechanisms.

Similarly, Selvakumar warned that blanket bans could be counterproductive, as users might turn to alternative platforms such as VPNs to bypass imposed restrictions.

“This not only reduces the effectiveness of regulation but also increases the risk of unsafe use, as users migrate to unmonitored platforms operating outside the domestic regulatory framework,” he said.

Both experts agreed that while temporary bans might be necessary to manage immediate risks, Malaysia’s long-term approach should be a flexible, risk-based regulatory framework that evolves with technological development to ensure user safety. - FMT