BANNER: Screenshot of an archived version of the removed deepfake video of Canadian Prime Minister Mark Carney. (Source: @smart.feed/archive) 

A deepfake video falsely depicting Canadian Prime Minister Mark Carney announcing controversial auto regulations originated on TikTok and spread to other social media platforms, including X, garnering millions of views. Even though the footage was labeled as AI-generated, debunked by fact-checkers, and eventually removed, the claims in the video continued to be amplified by influencers on TikTok, who further amplified the false information in the deepfake video to their followers. The altered video was first published directly before the 2025 federal elections and resurfaced after Carney was elected.  

Carney was elected as Prime Minister of Canada on April 28, 2025, following a snap election that took place amid an evolving information environment marred by many challenges, including annexation threats from the United States. During the electoral campaign, Canadians on Facebook and X were targeted with ads promoting scams disguised as political content. Additionally, the DFRLab identified bot-like accounts on X that flooded the official accounts of federal parties and their leaders with spam-like, political messages. And on election day, misleading content accusing Elections Canada of fraud went viral. The 2025 Canadian election was a showcase for many of the increasingly common threats to election integrity.  During the electoral campaign, Canadians on Facebook and X were targeted with ads promoting scams disguised as political content.

Moreover, malign actors have been exploiting increasingly accessible and easy-to-use AI tools to aid their activities, and Canada has not been immune to such threats. For instance, the previously mentioned scam campaign, promoted via social media ads, created inauthentic websites that mimicked legitimate Canadian media outlets, such as CBC. The actors behind the campaign applied deepfake technology to alter videos of Carney, which were added to the websites, falsely depicting him as promoting fake investment platforms. Since becoming the leader of the Liberal Party, Carney has been a frequent target of deepfake content. 

The viral deepfake TikTok video of Carney gained traction one day before the elections. The inauthentic video manipulated authentic footage, taken from a March 27 address in which Carney discussed Canada’s response to the latest US tariffs, to fabricate a video of the prime minister delivering a different message. In the deepfake video, Carney stated that the government would soon phase out older vehicle models to ensure “safety and emission standards,” along with other transportation regulations, possibly aiming to mislead users about the Canadian government’s long-term plan for zero-emission vehicles. The video depicted the Liberal Party leader discussing the acceleration of an existing government policy aimed at ensuring all new cars in the country are zero-emission by 2035. The timing of the video’s release appears to have contributed to its virality, as it emerged amid the election of a new leader, potentially increasing its ability to mislead users who would expect to hear about policy changes before or after elections. 

The original TikTok video was labeled as AI-generated, with a “Creator labeled as AI-generated” tag under the caption. The caption also noted that the audio was created by Fish Audio, a free AI text-to-speech audio cloning tool, and included a watermark of the software’s name on the video itself. The now-suspended account, @smart.feed, which previously posted videos created with similar AI tools, had also published different videos, some of which were clearly intended as parodies, showing Carney and Ontario Premier Doug Ford making exaggerated claims. Yet, the deepfake of Carney discussing transportation regulations appeared more deceptive and misleading as it imitated an official government announcement. AFP Canada, which documented and debunked the video, also previously debunked another AI-generated video by the same user. 

Despite the clear indicators of AI-generated content, the virality of the video appears to have been aided by the attention it garnered from US users, who responded with concern about such a policy being implemented in their neighboring country. These reactions from US users were likely exacerbated by the political tensions between the two countries.  

The DFRLab, at the time of research, noted that the now-removed video had garnered over two million views. An archived version of the post shows that it received over 37,000 likes and over 83,000 shares. A report on the video by Radio-Canada stated that it was viewed over three million times in one day. Moreover, comments on the TikTok video indicated that some users did not suspect that the content was fake. While the original video was removed, four posts with different versions of the deepfake video were still available on TikTok at the time of writing. 

While the originator of the Carney deepfake had posted several AI-generated videos to TikTok, this particular video is noteworthy for its cross-platform spread and ability to garner views on other platforms, showing the need for a cross-platform approach to combat the spread of deepfakes. A few days after the election, the video was repeatedly posted on X. AFP Canada also noted that the video found its way to other platforms such as Facebook and Instagram. On Meta’s platforms, although it is challenging to assess total reach, at least one video received nearly 19,000 views. On X, we identified eighteen posts that shared the video, with two of them receiving millions of views. One of these posts, was by @DisrespectedThe, was published on May 4 and received around 2.4 million views. A few hours later, another post by @LangmanVince received around 1.5 million views.  

Community notes were frequently added to X posts that shared the video, clarifying that the footage had been digitally altered and providing the original, unedited video source from Carney’s speech about tariffs. The post by @DisrespectedThe also received a “manipulated media” label alongside the community note, but other viral posts did not receive such a label. The inconsistent application of the “manipulated media” label to viral posts sharing the same misleading video footage raises questions about the effectiveness and uniformity of X’s content moderation policies. While some iterations of the video were appropriately flagged, others managed to evade the label, highlighting potential gaps in the platform’s detection and enforcement mechanisms. In contrast, the consistent application of community notes verification labels suggests a more robust system in place for user-generated fact-checking.  

Even though several news outlets and fact-checking organizations debunked the video, and community notes were quickly added to posts sharing it on X, the claims made in the video continued to be further amplified on TikTok by users with large followings who promoted the claims and attributed them to Carney. Such users commented and reposted the claimed statements without sharing the original video. This highlights how false information, once released into the information environment, can persist despite fact-checking efforts made against the originating source.  

One example of this was a video posted on May 5 by content creator, Grace Unfiltered, a user with almost 200,000 followers who regularly posts commentary about US news. The user posted a TikTok video in which he provided the same false information shared in the deepfake video, presenting it as truth and claiming it to be an official Canadian regulation. This video received almost four million views and tens of thousands of engagements, including over fifty thousand shares. Many comments on Grace Unfiltered’s video suggested that at least some users considered the information to be true. Similarly, the user Real White Rabbit also posted a video, repeating the claims, which received around 2.7 million views. 

Grace Unfiltered and Real White Rabbit’s cases are noteworthy, given that neither user is specifically focused on sharing Canadian news. TikTok’s engagement-driven algorithm, along with its popular casual front-facing commentary video format, enables such influencers to spread sensational and unverified claims that attract views without the need to add context or sources. While monitoring the spread of false information through the commentary-style videos poses certain difficulties, research by NewsGuard in 2022 highlighted the frequency of misleading claims in the platform’s search results and For You page. 

This case illustrates several key issues that arise when interacting with AI-generated content on social media platforms. Even when content is clearly labeled by the platform as AI-generated or manipulated, it can still deceive users and receive significant attention, especially when the content mimics decision-makers during and after critical political events like elections. More concerning, the fake claims made in such content can continue to gain traction and develop independently if amplified by popular users who may not provide clear context to their followers about the provenance of the claims, thereby extending the reach of these claims and making detection and assessment more difficult.

DFRLab, “Deepfake video of Canadian Prime Minister reachers millions on TikTok, X,” Digital Forensic Research Lab (DFRLab), June 19th, 2025,