Deepfakes of Real Influencers' Faces Fuel a Trade War Network, McGill Report Finds
McGill researchers found at least 293 TikTok and Instagram accounts using real creators' faces and AI voices to post 3,201 videos, most of them about Canada and the U.S.-Canada trade war.
A real creator's face. An AI voice. Words they never said.
At least 293 TikTok and Instagram accounts used real creators' faces and AI voices to post 3,201 videos, according to a report from McGill University's Centre for Media, Technology and Democracy. Most are about Canada and the U.S.-Canada trade war.[1]
What Happened
The report, "Stolen Faces," was published November 12, 2026.[1] The Canadian Press reported that the network is likely based outside of Canada and spreads "politically divisive" content.[3]
Researchers counted 257 TikTok and 36 Instagram accounts. Their videos, posted between August 7 and November 4, drew 22.1 million plays. All accounts were still active as of November 11. The report says its counts "likely underestimate the network's full reach."[1]
Tariffs, trade talks, boycotts and their fallout make up 54% of the videos and 71% of views. The network started during trade talks and a run of votes: Quebec's election, Alberta's referendum, British Columbia's election and the U.S. midterms.[1]
Of 187 videos about the Alberta referendum, 130 promoted separation. Researchers found "almost every account takes both sides."[1]
The most-played video claimed Carney answered a "48-hour ultimatum" from Trump "by cancelling $8 billion in U.S. defence contracts." The report says "This never happened." The story ran in 161 videos on 129 accounts and drew 5.1 million plays.[1]
Of the 293 accounts, 283 use footage of a single real person. Researchers counted 226 distinct people and identified 104. Only three TikTok accounts use fully AI-generated avatars.[1]
An AI text-to-speech tool supplies the voice. Another AI tool generates matching lip movements. A deepfake-speech detector rated the narration as AI-generated in 94% of videos. The report notes the "hair and eyes repeat exactly, but the mouth does not."[1]
Social media influencer Cole Bennett told CP24 that the latest deepfakes of him appeared last month, as he was posting about Alberta's referendum. "Some of my followers had started sending me videos," he said. "They were using an AI voice over my video to make it seem like I was saying something that I wasn't."[2]
Why It Matters
Platform labels missed most of these videos. TikTok's AI label appeared on 22 of 24 avatar videos, but on only 26 of 709 built on real people's footage.[1]
None of the 733 was disclosed as AI by its poster. Only 6.3% of commenters said or suggested a video was fake, AI-generated, a bot, copied or propaganda.[1]
Who runs the network is unknown. Posting and registration times point to East or Southeast Asia, and the report says the timing "rules out North America." Twelve of 184 checked accounts carry Chinese-language settings or are registered in China. The report cautions this "may reflect whoever supplied the accounts."[1]
The accounts "show no sign of earnings." "We do not know what the network is for," the report states.[1]
TikTok bars AI content made from a private individual's likeness without consent. Meta bars using someone's likeness "to impersonate them and deceive others."[1]
In Ottawa, the Strong and Free Elections Act received royal assent last June. Global Affairs Canada's foreign disinformation unit received more money. The Safe Social Media Act is at committee stage.[2] The report calls that bill "significant progress" and notes it requires labelling of synthetic content, but says it "will have limited impact" on preventing impersonation.[1]
What to Do
A familiar face is not proof of who is speaking. Check the account, not just the person on screen. A new account using a known creator's face is a warning sign.
Watch the mouth. In this network, the hair and eyes repeat while the lip movements are AI-generated.
Confirm big political claims with established news outlets before sharing.
Verify What You See and Hear
Upload any suspicious image or video frame at witness.vision/scan to check whether it was AI-generated. The Witness Chrome extension scans images directly on any webpage.
Scan now at witness.vision/scan
- [1]Centre for Media, Technology and Democracy, McGill University, November 12, 2026. https://mediatechdemocracy.com/en/publications/stolen-faces_2026/
- [2]CP24, Michael Serapio, November 16, 2026. https://www.cp24.com/news/canada/2026/11/16/ai-deepfakes-use-canadian-influencers-faces-to-spread-political-disinformation/
- [3]The Canadian Press via CTV News, November 12, 2026. https://www.ctvnews.ca/canada/article/a-deepfake-video-network-is-spreading-false-information-about-the-us-trade-war-report/