A McGill study found nearly 300 TikTok and Instagram accounts lip-syncing real creators’ faces to AI scripts about the trade war, and few viewers noticed.
Imagine scrolling past a video of yourself saying things you never said, to millions of strangers, about a political fight you may not even agree with. For hundreds of creators in Canada and the U.S., that is exactly what happened this fall.
According to The Globe and Mail‘s Murad Hemmadi, researchers at McGill University’s Centre for Media, Technology and Democracy found a network of 293 Instagram and TikTok accounts that “steals real people’s faces” to push content about politically touchy subjects to Canadians and Americans. The network posted more than 3,000 videos that drew upwards of 20 million views in roughly three months.
How the “stolen faces” worked
Some of the videos simply lifted popular creators’ existing clips, dropped them on new backgrounds and reposted them. Others kept the creators’ real faces but added what researchers judged to be AI-drawn lips reading AI-written scripts in AI-generated voices. “Words are being literally put in their mouths and shown to millions of people,” Aengus Bridgman, the centre’s associate director of research, told the Globe.
Hala Canada reported that the report, titled “Stolen Faces,” traces the operation back to Aug. 7, splits the accounts into 257 on TikTok and 36 on Instagram, and counts 3,201 videos with 22.1 million views. By the end of the monitoring period the network was posting about 100 videos a day. One example: a clip of American influencer Ashley Hetherington, originally about personal relationships, was reworked so she appeared to talk about Canadian job losses, U.S. tariffs and the Alberta referendum. Clips of finance creator Andrei Jikh were also altered.
The Globe said Canadian creators whose content or likeness was used included comedy and true crime creator Kallmekris, conservative commentator Jasmin Laine and mortgage broker Ron Butler. Edmonton political commentator Cole Bennett found his own get-out-the-vote message recycled with an AI voice and mouth on top of another creator’s footage. “What I say is the most important part of my brand and who I am online,” he told the paper, calling deepfakes like these “frightening.”
Fake claims, real views
Many of the most popular videos were flatly false. The most-watched script claimed President Donald Trump gave Canada 48 hours to drop its retaliatory tariffs and that Prime Minister Mark Carney answered by cancelling $8 billion in U.S. defence contracts. Neither happened, but the 161 videos using that script pulled in about 5.1 million views, per the Globe.
Viewers mostly bought it. Only 6.3 percent of comments suggested the videos might be fake or bot-made, and thousands of people replied as if they were talking to a real person. TikTok’s AI-labelling system flagged just 4 percent of the videos. “The network has found a loophole,” Bridgman said, explaining that because the deepfakes are built on real human footage, detection software reads them as real.
The researchers could not name who was behind it. The Globe reported that activity appeared to follow working hours in East or Southeast Asian time zones and some accounts were registered in China, though Hala Canada noted the report stresses that this does not prove any government was involved. The network did not appear to be pushing one ideology, and it did not yet seem to be making money directly from the videos.
Creators left to clean up
According to AFP’s report, published by The Straits Times, the real creators paid a price. Laine, whose content leans conservative, told AFP she found herself fending off accusations that she was being paid to repeat talking points she never said. “There you have your face and your voice and what looks like you speaking, in some cases, to something that you actually don’t stand for,” she said. Bridgman called those accusations a “very good example of harm,” adding that viewers “have fallen hook, line and sinker for the network content.”
Butler told AFP this was not the first time he had been impersonated, and he sounded resigned. “You can’t stop this because there’s too many people who can do it,” he said. “It’s too easy to do.”
A TikTok spokesperson told AFP the company permanently banned the accounts named in the report, plus others, for breaking its rules on misleading content. Meta declined to comment to AFP but said it had taken down the Instagram accounts involved. AFP also noted the researchers’ 4 percent figure came before TikTok launched a new deepfake detection tool.
Takeaway
Your face and voice are your business, and this case shows how cheaply both can be hijacked to say anything. Creators should search the major apps for reposts of their own clips now and then, report impersonators fast, and give followers a simple way to check what’s really from them, like a pinned post listing official accounts. Platforms’ AI labels clearly aren’t catching this kind of fake yet, so the first person to spot it will often be you or your fans.
Image: AI-generated illustration.
