They have perfect hair, flawless makeup, and absolutely nothing to say that is true. A CNA investigation has uncovered a coordinated network of AI-generated female TikTok presenters pumping out disinformation about Singapore and Malaysia, racking up more than 3 million views in the process. The whole operation looks like a propaganda factory running on recycled scripts, stolen voices, and the reasonable assumption that nobody is paying close enough attention.
The Factory Floor
CNA's investigation examined 30 TikTok accounts and found more than 550 videos behind them. Almost all of them, 98 per cent, were built the same way: take an AI-generated or manipulated female persona, attach a reused voiceover, slap on a recycled script, and post. Rinse. Repeat. Nearly nine in ten of those videos pushed false or misleading claims.
The accounts ran between October 2025 and June 2026, churning out content in Mandarin about geopolitics, economics, and foreign relations. The presenters looked polished and credible. They spoke with authority. They were, in the most literal sense possible, made up.
When CNA sent TikTok questions and flagged two specific accounts as examples, the platform shut both down within days. A spokesperson confirmed the accounts had violated rules on "deceptive behaviour," which TikTok defines as covering covert influence operations, impersonation, spam, and fake reviews. Good. Now what about the other 28 accounts in the investigation, and whatever else is still out there that nobody has flagged yet?
The Lies Were Good Enough to Fool Three Million People
Here is the part that should genuinely alarm you. These videos did not lead with wild, easily dismissed conspiracy theories. They led with real facts. Singapore's port statistics. ExxonMobil's plant closure. China's status as a major trade partner. True things, presented accurately, used as on-ramps to complete nonsense.
One video claimed Singapore's port dominance was on the verge of collapse because China's Hainan Free Trade Port had gone fully operational. Singapore's port actually handled a record 44.66 million containers in 2025 and has ranked as the world's second-busiest container port for fifteen straight years. The collapse, it turns out, was not happening.
The most brazen fabrication involved Singapore's Foreign Minister Vivian Balakrishnan. A claim spread across these accounts alleged that he had unsuccessfully begged China and Indonesia not to let a new shipping route bypass Singapore. That event did not occur. It was invented wholesale. That specific invented story was repeated across different accounts over nearly two months and viewed more than 100,000 times.
Fake Consensus Is the Product Being Sold
What makes this operation genuinely clever, and genuinely dangerous, is the structure of the illusion. CNA found 94 videos consistently repeating false or misleading claims, accounting for over 1.6 million views on their own. Across 24 accounts, different AI-generated "presenters" took turns delivering the same talking points over weeks and months.
To anyone watching casually, it looked like multiple independent commentators were all arriving at the same conclusions about Singapore and China. That is not commentary. That is manufactured consensus, and TikTok's format is practically designed for it. Short videos, high volume, algorithm-driven repetition. You see the same narrative six times from six different faces, and your brain quietly files it under "things people seem to agree on."
Associate Professor Saifuddin Ahmed from Nanyang Technological University told CNA that the goal appears to be eroding trust and fracturing social cohesion. "Any audience that grows to distrust its own government's narrative is more susceptible to external influence over time," he said. Which is, of course, the whole point.
Who Is the Target, Exactly?
The Singapore-focused content consistently pushed a worldview where Singapore's prosperity depended on its alignment with China. The Malaysia-focused content went a different, uglier direction, with videos blaming Malaysia's development on policies that allegedly sidelined its ethnic Chinese population. Two countries, two different emotional levers, same coordinated operation.
Experts who spoke with CNA suggested the audience may extend well beyond Singapore and Malaysia themselves. The global Mandarin-speaking diaspora is a plausible target. Assoc Prof Saifuddin pointed out there is real strategic value in shaping how international investors perceive the political stability and orientation of these countries. Disinformation is not just about changing minds. Sometimes it is about creating enough noise and uncertainty that confidence erodes on its own.
A previous CNA investigation in February found similar narratives being pushed on YouTube. Experts told CNA they have observed these same campaigns running on other platforms too. This is not a TikTok problem. TikTok is just where this particular investigation looked.
This Is Not a New Problem. It Is Getting Harder to See.
The uncomfortable reality is that nothing about this operation required extraordinary technical sophistication. AI-generated presenters are not some cutting-edge classified capability anymore. The tools to build a convincing fake female news anchor and attach a voiceover to her are widely available, cheap, and getting better every month.
What this investigation actually demonstrates is a scaling problem. A human running a disinformation operation can only produce so much content. An AI-assisted operation can produce 550 videos across 30 accounts and generate 3 million views while the operator is asleep. The volume is the point. The repetition is the mechanism. And the realistic-looking woman on screen is just the packaging.
TikTok removed the accounts CNA flagged. That is the good news, such as it is. The bad news is that the barrier to spinning up new accounts is approximately zero.
The Dingo Take
Let us be clear about what CNA documented here. Someone built a small factory to manufacture fake credibility, staffed it with AI women who do not exist, gave them recycled scripts full of lies dressed up in real statistics, and aimed the whole thing at millions of people who had no particular reason to be suspicious. It worked. Three million views. One invented anecdote about a foreign minister begging for mercy, watched over 100,000 times. That is not a glitch in the system. That is the system being used exactly as an influence operation would want to use it.
The platforms keep saying they have rules against this. TikTok told CNA it "constantly monitors" for coordinated manipulation. And yet CNA, a news organisation, had to do the detecting, flag two specific accounts, and wait a few days for action. Meanwhile the other 28 accounts in the investigation sat there. The content is already out there. The 3 million views already happened. You cannot unwatch a lie once it has been filed away in someone's brain as "something people seem to agree on."
The wider lesson here is one that keeps arriving from every direction and keeps getting ignored. AI does not create disinformation problems from scratch. It turbocharges the ones that already exist, removes the labor costs, and makes scale trivially easy. Someone decided Singapore needed to look like a Chinese client state. Someone decided Malaysia's ethnic politics were a useful wedge. Then they let the machines do the talking. We should probably take that seriously before the next investigation finds 5 million views instead of 3.
