SHAH ALAM - Malaysia must prepare now for AI-driven manipulation ahead of the 16th General Election (GE16), as fake accounts and tailored political messaging can shape public perception without voters realising, a security and political analyst has warned.
Universiti Teknologi Mara (UiTM) psychological warfare expert Dr Noor Nirwandy Mat Noordin said the use of AI and profiling to target political issues was becoming increasingly common, but that did not make it acceptable.
“Something that has become a habit is not necessarily good,” he said in an interview with Sinar Harian.
He was commenting on Anthropic’s disclosure of about 1,000 fake accounts on X targeting Malaysian users, with profiling based on all 222 parliamentary constituencies.
‘Weaponisation of AI’
Nirwandy said generative AI could produce voice and video content, while agent-based AI could be programmed to carry out tasks and spread content automatically. In politics, he linked this to the “weaponisation of AI”, where the technology is used as a tool to achieve specific ends.
He said the possibility of such manipulation was real and the country needed to understand whether the technology was being used for political purposes, what the aim was and who was involved.
On digital profiling, he said it allowed information to be organised according to the traits and leanings of certain groups, so people needed to understand how their data was stored, who controlled it and who could access it.
Fake engagement
Nirwandy said likes, shares and interactions could be bought to create the impression that content was widely supported.
“An issue may appear to have broad support when the interactions displayed do not necessarily reflect the real views of the public,” he said.
He added that “humanisation” techniques could make accounts, comments and discussions look as if they were run by real people.
He said it was difficult to tell bots, fake accounts, human-run accounts and AI-assisted accounts apart because the categories overlapped. Human-run accounts could also behave in a non-organic way when their activity was coordinated.
“Assessment cannot be based on a single indicator. We need to look at the overall pattern and the context of the content,” he said.
Half-truths harder to counter
Nirwandy said algorithms made modern information warfare more dangerous than traditional propaganda because content could be delivered repeatedly and tailored to users’ leanings, shaping perception gradually.
He distinguished between misinformation (false information spread without necessarily intending to deceive), disinformation (deliberately spread false or misleading information) and malinformation (genuine information used in a harmful context).
True facts taken out of context could create a misleading impression, he said, so people should judge not only whether information was true but also its context, purpose and delivery.
Regulation and detection
He said Malaysia had a regulatory framework, including the Communications and Multimedia Act 1998 (Act 588) and a licensing mechanism for internet messaging and social media service providers that meet set criteria. However, licensing was not an absolute guarantee against digital threats.
“We still need a combination of regulation, technical security standards and the responsibility of platform providers,” he said.
He noted that Malaysia was largely a user of social media platforms rather than their owner, which limited its ability to detect fake networks on its own.
He said detection capability should be strengthened through cooperation between agencies, compliance by platform providers and technology development and suggested the country consider building its own digital platforms, including through regional cooperation.
If a fake video of a candidate surfaced 24 or 48 hours before polling, he said the public could report it to the authorities, including the Malaysian Communications and Multimedia Commission (MCMC). Public awareness remained the most important safeguard.
Young people and extremism
Nirwandy also warned that social media and online games could expose young people to extremist ideas. He cited nihilism, the loss of belief in values and meaning, as a mindset that could lead to radicalisation.
He said individuals could be manipulated in stages, for example by being invited into a small group in an online game before being drawn into closed chat communities. Language patterns and shared posts could be indicators, but they should not be treated as sole proof of extremism.
He called for discussions in schools, communities and mainstream media to build “cognitive resistance” among young people so they could assess, question and reject attempts at manipulation.
Three checks
Asked what he would check first if an issue were backed by hundreds of seemingly ordinary accounts, Noor Nirwandy named three things. First, the objective of the campaign, and whether it serves the public or narrow interests. Second, the accounts involved, and whether they represent real individuals. Third, the digital profiling and messaging, and whether it reflects real human views or AI-generated content.
He recommended a “pause, relax, observe” approach: re-evaluate the information received and understand why the message was sent.
He added there was no evidence that every possible threat would materialise, but mental defences and national security had to be strengthened through a whole-of-society and whole-of-government approach.