‘Technology embedded in broader digital ecosystems than standalone platforms’
PETALING JAYA: The Malaysian Cyber Consumer Association (MCCA) has said regulating generative AI platforms poses far greater challenges compared with blocking individual websites.
It was commenting on whether similar measures could be applied to AI video-generation tools following a recent move by the Malaysian Communications and Multimedia Commission (MCMC) to block access to the websites of dating platforms Grindr and Blued.
MCCA president Sirajuddin Jalil said the rationale behind blocking such dating platforms differs from regulating generative AI.
“Those dating apps are not just about content, they are also often seen as conflicting with local norms and values.
“The second reason is pornography is totally banned in Malaysia, and within those platforms there may be pornographic-type material and other content that could be against the Communications and Multimedia Act 1998.”
However, he said regulating AI platforms is more complicated due to how the technology operates across multiple digital services.
“When we talk with the government, anything that is favourable to the government and the public, they will consider it.
“The real issue is: what is the bottleneck for MCMC or the government to block AI-related websites? Take ChatGPT and Gemini for example, content generation can happen there,” he added.
He said many AI tools are embedded across broader digital ecosystems rather than operating as standalone platforms.
“Gemini is everywhere. It is the same with WhatsApp and Meta,” he said, explaining that it is difficult to ban a particular technology that is used by several platforms.
“It is not easy because it is embedded in many other products. How do you ban Meta AI when Meta is inside WhatsApp?”
Sirajuddin said many AI tools are built on underlying systems known as large language models (LLM).
“AI is not like an app. AI is the algorithm and the machine learning system, which is LLM.
“People use the LLM app programming interfaces (tools that allow apps and AI systems to connect and work together) to create another product, and another product.”
He added that the global ownership of many AI models complicates enforcement.
“We go to ChatGPT. We go to Gemini. That is the vulnerability. All of it is not owned by Malaysians. Malaysians are not using Malaysian LLM,” he said.
“Higgsfield (a website for AI-generated videos) has many engines inside. It has Seedance (ByteDance’s AI video model), it has Kling (a China-based AI video generation model), it has other video engines, it has Veo (Google’s AI video model)… it combines them.
“So, it is similar to others. They do not create their own model because to create one LLM costs millions and it is extremely expensive.”
He said reliance on foreign AI systems raises questions about how effectively domestic regulators can control such technologies.
“The National AI Office is supposed to be a policy house for AI.
“If it is a policy house for AI, then it must consistently produce communication related to AI policy to the public.”
Sirajuddin also said regulators must balance the risks of misuse with the technology’s wider benefits.
“The reality is those tools are also used for many developments that bring benefits.
“Before banning them, if the government does not have a clear understanding and does not call AI operators to discuss what the problem is and ask whether they are aware of it, then it becomes a ‘bottleneck’.”
The debate comes as deepfake videos involving public figures have surfaced online, including AI-generated TikTok clips with fake images of the King of Malaysia Sultan Ibrahim, that prompted a warning from the Johor palace on Feb 25 and a police probe the following day.
On Friday, a widely spread video involving Border Control and Protection Agency Director-General Datuk Seri Mohd Shuhaily Mohd Zain was confirmed by authorities to have been manipulated by deepfake technology.
Communications Deputy Minister Teo Nie Ching told the Dewan Rakyat on Monday that 90% of 115,161 requests submitted to social media platforms to remove online content were complied within the first two months of this year.
The Sun Malaysia
