PETALING JAYA: A smartphone and a 15-second audio clip are all it takes to produce a convincing fake video of an election candidate, says the Malaysia Cyber Consumer Association (MCCA), as videos of politicians made using artificial intelligence (AI) spread across social media.
Its president Siraj Jalil said the technology had become "dangerously accessible".
"Anyone with a smartphone or basic laptop, a 15 to 30 second audio clip from a candidate's public speech and off-the-shelf generative AI tools can clone a voice and sync lip movements in minutes, for free or just a few ringgit," he said.
Such videos have mushroomed on all platforms recently, many of them satirical clips that show politicians saying funny things they never said.
Others carry false claims. During the Melaka state election campaign, Barisan Nasional deputy chairman Datuk Seri Mohamad Hasan warned voters about a fake TikTok video made with AI that showed him promising RM25,000 to everyone.
Communications Minister Datuk Fahmi Fadzil reminded voters on Thursday to be careful with content that may be AI-generated, but said the Malaysian Communications and Multimedia Commission (MCMC) had not forwarded any complaints to him so far.
ALSO READ: M'sia acts on AI election threat
Siraj said politicians were easy targets because their livestreams and press conferences gave bad actors plenty of high-resolution material to work with.
To spot a deepfake, he said voters should watch the mouth, teeth and jawline for slight blurring, unnatural blinking or teeth that look too smooth during fast speech.
They should also listen for a flat or overly rhythmic voice, especially in local Malay accents, and for abrupt cuts in background noise.
The source is another giveaway, he said, as deepfakes almost always come from newly created, anonymous or unverified pages instead of official candidate channels or established news outlets.
"Any shocking statement not reported by mainstream media should be treated as fabricated until verified," he said.
Complaints to MCMC over deepfake content rose from 917 in 2024 to 3,612 last year and 7,967 as of June 15 this year, Deputy Communications Minister Teo Nie Ching told the Dewan Rakyat in June.
In a written parliamentary reply last month, Fahmi said platforms had removed 713 deepfake posts involving political leaders between Jan 1 last year and Sept 15 this year at MCMC's request.
Siraj said social media platforms were not removing such videos fast enough as moderation was still largely reactive.
"During a compressed election campaign, speed is everything," he said.
He said a video could take hours or even days to be flagged, reviewed and removed, by which time it would have been screen-recorded and shared in closed WhatsApp and Telegram groups that platform moderation cannot reach.
"A delayed takedown on Facebook does little to undo the damage," he added.
Siraj noted that MCMC had introduced the Risk Mitigation Code under the Online Safety Act 2025, which requires licensed platforms to label AI-generated content.
The code, in force since June 1, covers manipulated or AI-generated content that resembles real people or events, and requires platforms to factor periods of heightened risk such as elections into their risk assessments.
Non-compliance can lead to financial penalties of up to RM10mil.
But Siraj said MCMC must hold platforms to stricter and more proactive standards before the 16th General Election, which must be held by February 2028.
These include automated detection and visible watermarking at the point of upload, instead of relying on users to declare that their content is AI-generated.
He also called for election "war rooms" at the platforms, with a strict deadline of under two hours to take down verified election deepfakes flagged by the Election Commission, candidates or accredited fact-checkers.
Paid boosting and algorithmic amplification of political content that lacks verified provenance should be banned as well, he added.
