That's impressive.
Thank you, Mr. Chair.
Two days ago, our researchers at the Canadian Digital Media Research Network published a report on a network of 20 YouTube channels—with nearly 40 million views—targeting Albertan audiences over the past year. The channels use AI-generated avatars of Premier Smith and Prime Minister Carney, paid American voice actors reading templated scripts and maps placing the western provinces as the 51st through 54th states of the United States. Real Albertan grievances are being co-opted and amplified by inauthentic infrastructure at scale six months from a provincial referendum on secession.
Civic resilience in 2026 is not an abstraction. Frontier AI models are advancing on a cadence driven by hundreds of billions of dollars of private investment, concentrated political endorsement across major powers and genuine breakthroughs in capabilities. What follows will almost certainly be a period of rapid social transformation and real upheaval, including labour market shocks, epistemic disruption, reorganized public services and new patterns of political organization and contestation.
What's more, the institutions we've entrusted with our democratic values are not well suited for this moment. Parliaments, courts, regulators, public broadcasters, universities and the press were designed for slower, more deliberative politics. Their legitimacy rests on judgment exercised in public and held to account over time. This cadence is a feature, not a bug, of our democratic system, but it is being asked to hold against a rate of change it was never built for.
Trust has eroded over a decade, in part because these institutions were slow to adapt as the information environment was being remade. They enter this transition already weakened and facing a compounding risk. This weakness is not incidental but is too often sought both by foreign adversaries working to accelerate our institutional decline and by a domestic political current that runs from openly anti-democratic movements to a quieter indifference about whether traditional institutions survive.
Our research has documented three shifts in the information environment that bear directly on the challenge we're about to face.
The first is social fracturing. We've long understood information fragmentation as something that happens within platforms: filter bubbles on Facebook and echo chambers on X. However, when you look at the information environment as one system, which we now do, by tracking coordinated signals across YouTube, X, Instagram, Facebook, TikTok, Reddit, Substack and all podcasts, what you see is different. Canadians are being fractured as a country. The Alberta network is one example of how the fracture is produced and exploited. When social cohesion erodes at a moment of open hostility from the United States, so does our sovereignty.
The second is who intermediates information for Canadians. Our work has shown that a majority of political news now reaches Canadians through influencers rather than journalists, and AI systems themselves are now becoming news intermediaries. Our audit of four major AI systems proved that they were trained on Canadian journalism, substituted their own responses for it and rarely credited the newsrooms that produced it. Information integrity was anchored in institutions with editorial standards, libel exposure and professional norms. It's now distributed across individuals and technological systems with none of these yet developed.
The third is AI-native manipulation at scale. The Alberta network produced 12 times more pro-annexation content than all authentic Albertan YouTube channels combined. The scripts were templated, the voices synthesized or paid for and the avatars fabricated. These are not sophisticated operations, though, because they don't need to be. Generative AI has driven the cost of producing such content to near zero, while platforms continue to carry it in Canada to Canadian audiences at scale. The origin and intent of these operations remain unresolved because YouTube holds the data.
In my view, then, civic resilience requires two things built together.
First, we need a regulatory regime that compels platform and AI co-operation with accredited researchers. The legislative vehicles already exist for this: a reintroduced online harms act with strong researcher access provisions, extended to cover consumer-facing AI products.
Second, we need the independent research infrastructure to make those provisions work. Canada is defunding this infrastructure at the moment it most needs it, as my colleague, Aengus Bridgman, testified to this committee last week. As the Hogue commission's 48 recommendations called for, we urgently need a standing national capacity, at arm's length from platforms and government, with structural funding rather than departmental line items that are vulnerable to budget cycles.
Wide-ranging efforts are needed to build civic resilience, as this committee has heard, but none of them will be possible if we don't understand and ensure the integrity of the information environment on which our democracy depends.
Thank you. I look forward to talking about it.