PETALING JAYA: Fake videos made with artificial intelligence (AI) are unlikely to decide an election on their own, but they could tip the scales in marginal seats where only a few hundred votes separate winner and loser, say analysts.
They said such videos need not win over voters in large numbers, as planting doubt or keeping supporters at home could be enough in a tight race.
Universiti Malaya political analyst Dr Mohammad Tawfik Yaakub said the danger was greatest late in a campaign, when candidates have very little time to respond while misinformation spreads rapidly on Facebook and WhatsApp.
Melaka goes to the polls on Saturday (Nov 14), with campaigning ending at 11.59pm on Friday.
"In a closely contested seat, a fake video does not necessarily need to convince thousands of voters. Influencing a few hundred voters or discouraging them from turning out could be enough to change the outcome," he said.
Tawfik said the most vulnerable were undecided voters and those with limited digital literacy, who may struggle to tell genuine content from AI-generated material.
"However, even educated and digitally literate voters can be misled if the video reinforces their existing political beliefs," he said.
While the opponents of the targeted candidate usually gain at first, he said the tactic could backfire badly once the video is exposed as fake.
"Voters may sympathise with the victim and perceive the opposing side as desperate, unethical and willing to win at any cost," he said, adding that this would hurt most among young voters and fence sitters who value integrity and transparency.
He cautioned, however, against assuming that the party which benefits is the one behind the video, as it could also come from anonymous supporters or third parties.
Universiti Tunku Abdul Rahman political analyst Dr Liew Wui Chern said one video alone may not change an election, but even a small effect can matter in a very close seat.
"The bigger risk is not always changing a voter's choice. It can create doubt, reinforce negative views or discourage supporters from turning out," he said.
Voters who rely heavily on TikTok, Facebook, WhatsApp or short videos are more exposed, he said, especially if they do not check the original source.
Liew also urged caution before calling such videos organised cybertrooper work, as they are now so cheap and easy to make that ordinary supporters can produce them without party machinery.
"But if many accounts push the same message at the same time, then coordination becomes more likely," he said.
He said the current framework was probably too reactive for an AI driven campaign, as the Malaysian Communications and Multimedia Commission (MCMC) generally investigates campaign content only when complaints are received.
"A fabricated video can accumulate thousands of views before a complaint is lodged, verified and acted upon. In a campaign lasting only two weeks, a correction arriving 24 or 48 hours later may already be too late," he said.
Before the 16th General Election (GE16), Liew said the Election Commission and MCMC should have a dedicated rapid response mechanism for synthetic political content.
This should cover faster verification, direct escalation channels with major platforms, preserving evidence from disappearing accounts and swift public clarification when a candidate has clearly been impersonated.
He added that regulation must still distinguish deceptive content from legitimate political satire or campaigning.
"The objective should be to stop impersonation and fabricated statements, not to police ordinary political argument," he said.
Meanwhile, lawyer Ang Woei Shang said the use of AI itself should not be the central issue, as it was merely a technological tool like conventional video editing software.
"A well produced AI video cannot turn a baseless accusation into a credible argument. Equally, the use of AI does not automatically invalidate a legitimate criticism supported by facts," he said.
He said the public had also grown accustomed to such videos on social media, with many viewers treating them as entertainment, political satire or humorous commentary instead of taking every scene literally.
"AI may determine how effectively a message is presented, but it is the substance of that message that ultimately determines whether it persuades the public or backfires on those who promote it," he said.
