OTTAWA - Social media influencer Cole Bennett has built an impressive following over the past two years, attracting hundreds of thousands of followers with his political commentary on Alberta, Canada and the United States.
It is a self-made career many 29-year-olds might envy. But it has also come with an unwelcome downside: artificial intelligence deepfakes that steal his image and words, repurposing them for content he never created or endorsed.
The latest incidents happened last month, as Bennett was posting about Alberta’s referendum.
“Some of my followers had started sending me videos,” he said. “Some videos would use my face, but the words coming out of my mouth weren’t my words. They were using an AI voice over my video to make it seem like I was saying something that I wasn’t.”
And Bennett is far from alone.
A new study by McGill University’s Centre for Media, Technology and Democracy has identified a network of at least 293 TikTok and Instagram accounts using deepfakes to spread politically divisive content.
The accounts take the faces of real social media influencers, overlay them with AI-generated voices and manipulate the footage to make it appear as though those individuals are saying things they never said.
What’s worse is that over a 3-month period, some 3,000 videos from these accounts were viewed over 20 million times.
“We were leading up to a referendum and a vote, so people are online,” Bennett said. “They’re researching, and these are the videos that are going up in their algorithm.”
And it wasn’t just the Alberta referendum.
The McGill study also found deepfake videos pushed content on the B.C. and Quebec elections, and the U.S. midterms.
Researchers could not determine who was behind this network. At best, they determined it was run outside of North America, possibly in east or south Asia.
But even without knowing the network’s motive, researchers did conclude, “its operators are willing to steal faces and spread falsehoods that set Canadians against the United States, against their own government, and against each other.”
It’s a conclusion that raises alarms about AI-driven content, and its potential to shape public opinion and undermine Canada’s democratic process with disinformation.
The federal government has taken several steps to combat deepfakes and disinformation campaigns.
They include the Strong and Free Elections Act that received royal assent last June, more money for a unit of Global Affairs Canada that exposes foreign disinformation campaigns, and the Safe Social Media Act which is currently at the committee stage in Ottawa.
But despite all these measures, tech experts warn it is not enough, as even social media companies don’t have the capacity to police the sheer volume of content that their own apps help generate.
“It’s so easy to do,” said Mohit Rajhans, a media and tech expert. “Whether you’re 15 years old or you’re 55 years old, you don’t even have to have a real agenda. You can do it out of pure boredom, and you can actually create more chaos and noise and conversation by doing something just because you can do it, rather than actually have an agenda anymore.”
Rajhans points out, part of the problem lies in the business model of social media companies. The same technology that creates AI for legitimate ventures is the technology used to create more harmful content, and as long as that technology helps generate income for the one, it’s hard to dismantle or hinder it to address the other.
In the short term, Rajhans says governments and social media companies need to work together more closely, sharing information and communicating at a level that has not yet been seen.
“If you can actually put the surveillance technology around some of the stuff that is happening, you will start to find the people that you can hold accountable,” he said.
In the longer term, Rajhans says the focus will have to be on continuous education, noting that technology is evolving faster than governments can introduce and implement regulations.
“We have to revisit whatever technology solutions that we have way more consistently in order to feel secure about what’s happening around us.”


