AI chatbots are a new front in the fight against disinformation

Sinead Bovell is a futurist and strategic foresight adviser to governments, security agencies, multilateral organizations and Fortune 500 companies.

During the self-described “freedom convoy” protests in Canada in 2022, a curious player emerged to amplify the grievances of the demonstrators: Russia. Russian state outlet RT covered the convoy far more than any other international media organization, despite the fact that Canadian vaccine mandates had little direct impact on Russian citizens. The protest mattered to Moscow less as a on public health policy than as raw material. It fit neatly into the Kremlin’s now-familiar pattern of exploiting polarizing events in Western democracies in order to deepen distrust in public institutions and undermine national cohesion.

The goal of an influence operation is to shape how a target audience thinks, decides or acts in ways that serve a foreign actor’s interests. Digital networks have become a central battleground for these campaigns, and today’s state-backed operations routinely use everything from social-media bots to sympathetic foreign outlets. Canadian intelligence agencies have responded with dedicated units built to both defend against, and operate within, this contested information environment. But artificial intelligence is to open an entirely new front.

According to a 2026 study by Oxford University and Reuters, approximately 10 per cent of people globally already turn to AI chatbots for news. That number rises to 17 per cent for those aged 18 to 24 years old. As AI is built directly into messaging apps, engines and personal devices, those figures are only going to climb. Much of what these chatbots both learn and pull answers from is the internet. This could be why the Kremlin-linked Pravda network has been flooding the web with pro-Russian material that leading chatbots have been documented repeating . It’s a tactic known as large language model grooming.

Opinion: AI companies must be held responsible for the hallucinations that bog down courts

Therefore, the target of influence operations in this emerging AI information environment may no longer just be people, but also AI models. Unlike social-media influence, which leaves a public record that researchers can in theory investigate and counter, exchanges between people and AI assistants are largely private. By the time someone enters a public debate, their understanding of the issue may already have been shaped through conversations with an AI chatbot.

Canada has to think critically the AI models Canadians will trust to explain the world, and which parties may have a vested interest in influencing them. As it stands today, none of the AI models Canadians routinely engage with are domestically owned. But reliance on foreign models need not mean surrendering a say in how these systems shape public life. Canada can require AI systems to rely on verified sources for high-stakes matters such as voting and banking, and form alliances with like-minded allies that strengthen its leverage in setting the standards for areas such as identity, provenance and accountability in machine-to-machine interactions. Intelligence agencies will also need the technical capacity to operate in and defend this environment, ensuring Canada’s public institutions remain accessible as people increasingly seek information through AI systems.

Canadians have a role to play in information defence, starting with recognizing that AI systems are never neutral and making a habit of asking chatbots for sources, verifying those sources, and seeking out other points of view. Canada can also learn from Sweden’s model of public resilience, which includes a Psychological Defence Agency that trains civilians to detect malign influence operations and makes clear that everyone living in the country is part of that defence.

Opinion: The menace of the machine that’s ‘happy to help’

The last major shift in how people get information offers a warning what happens when institutions wait too long to shape how the environment unfolds. The extent to which social media and digital platforms have been exploited, and the toll on democratic discourse and social cohesion, are now clear. The transformation is under way, and Canada would be wise not to let it unfold the same way.

Four years after the self-described “freedom convoy,” foreign actors are again surfacing around a Canadian political fault line, this time Alberta separatism. Intelligence officials and researchers have warned Canadians to remain vigilant the narratives circulating online.

Less clear is what narratives are forming in the conversations Canadians are having with AI assistants.