August 3 Bot Activity on X Related to Rally in Kirkland Lake

Résumé analytique

On July 31, 2024, Conservative Leader Pierre Poilievre held a rally in Kirkland Lake as part of a tour of Northern Ontario. Three days later, on August 3, hundreds of X accounts posted about the event, claiming to have attended and using language like "buzzing with energy" and "as a northern Ontarian." The timing, character of the messages, and features of the accounts raised widespread speculation about their origin, including whether a political party or a foreign entity was responsible. The Canadian Digital Media Research Network investigated the event through six rapid incident updates, concluded with this debrief.

The investigation found this minor incident was likely caused by a single entity or actor using newly created bot accounts with posts composed by a low-quality or poorly prompted generative Large Language Model. This bot network consistently posts about recent news topics and is only incidentally interested in Canada or Canadian politics. Very few Canadians saw the original posts, so the direct impact was negligible, but subsequent discussion of the incident garnered millions of views on X and likely millions more through amplification by traditional Canadian media, with an estimated several million Canadians hearing about the incident. Nearly half of Canadians who heard of the event believe a political party or partisan individuals were responsible, and of those, most blamed the Conservative Party — but the investigation found no evidence that a political party or foreign entity employed this bot network for political purposes.

The incident highlights three vulnerabilities: current technology allows cheap bot accounts combined with accessible generative AI to scale information operations rapidly with few guardrails; a lack of cooperation and transparency from platforms (X provided no information on measures taken, and 98% of the 300 accounts captured in screenshots were deleted, suspended, or banned) leaves researchers and journalists with limited ability to respond; and the way Canadian media and politics discuss information operations, dominated by partisan blaming without evidence, worsens the problem. Taken together, these vulnerabilities could enable more sophisticated future efforts, such as targeted harassment, discrediting public figures via deepfakes, hijacking key democratic moments, impersonating trusted sources, and eroding public trust in the information environment.