Fact-based journalism that sparks the Canadian conversation

Articles Business Environment Health Politics Arts & Culture Society

Special Series Hope You’re Well For the Love of the Game Living Rooms In Other Worlds: A Space Exploration Terra Cognita More special series >

For the Love of the Game

In Other Worlds: A Space Exploration

More special series >

Events The Walrus Talks The Walrus Video Room The Walrus Leadership Roundtables The Walrus Leadership Forums Article Club

The Walrus Video Room

The Walrus Leadership Roundtables

The Walrus Leadership Forums

Subscribe Renew your subscription Change your address Magazine Issues Newsletters Podcasts

Renew your subscription

The Walrus Lab Hire The Walrus Lab Amazon First Novel Award

Amazon First Novel Award

Alberta Separatism Fuelled by Russian Networks and US Influencers

A sweeping new report traces how online campaigns are amplifying division inside Canada

Canada is facing the kind of political warfare most citizens were never taught to recognize. It doesn’t look like tanks at the border. It doesn’t arrive with a formal declaration. It appears as viral videos, anonymous websites, influencer commentary, algorithmic outrage, fake local voices, leaked voter data, and foreign actors amplifying domestic grievances until citizens can no longer tell where their own debate ends and someone else’s operation begins.

Another term for this is cognitive sovereignty, or the ability of citizens to make political decisions freely, without foreign manipulation, coercion, or engineered distortion. It is not a media studies concept. It is a national security concept. It describes something citizens possess the way they possess physical safety, the way a country possesses its territory, the way a democracy possesses its institutions. And like all of those things, it can be taken from you.

Alberta is now the Canadian test case.

On May 6, 2026, a team of Canadian researchers released one of the clearest public descriptions yet of the threat. The report, Decision Making & National Unity Under Threat: Foreign Interference, Cognitive Sovereignty, and the Alberta Referendum, was published by DisinfoWatch, the Global Centre for Democratic Resilience, the Canadian Digital Media Research Network, CASiLabs, and the Media Ecosystem Observatory.

The report’s central finding is stark: Canada’s cognitive sovereignty is being actively targeted by foreign governments, state-aligned media networks, ideological operatives, and profit-driven manipulation systems, all of them pressing on the same fault line at the same time. That fault line is Alberta.

This is not because Alberta’s grievances are illegitimate. They are not. Albertans have real and long-standing concerns about energy policy, federal–provincial relations, and economic fairness within Confederation. Those concerns deserve to be debated openly and democratically by Canadians, among themselves, on their own terms. The danger is who has joined the debate and why.

Before we can understand what is happening in Alberta, we need to understand the technology that makes the assault on cognitive sovereignty possible at the scale we are now seeing.

In October 2025, Stanford researchers Batu El and James Zou published “Moloch’s Bargain: Emergent Misalignment When LLMs Compete for Audiences” that demonstrated something the artificial intelligence industry had suspected but not yet fully quantified. When AI systems are optimized to win in competitive environments, including elections, they produce deception as a predictable output, regardless of any instruction to the contrary.

The shocking findings: to win roughly five more votes out of every 100, the AI systems produced nearly five times more disinformation. To get a modest boost in social media engagement, they generated almost twice as much false content; not twice as much total content but twice as much false content specifically.

Every gain in competitive performance came packaged with a disproportionate rise in deception. (For those who want the specifics: a 4.9 percent gain in vote share coincided with a 22.3 percent increase in disinformation and a 12.5 percent increase in populist rhetoric; a 7.5 percent social media engagement boost came with a 188.6 percent increase in disinformation; a 6.3 percent sales lift in marketing accompanied a 14 percent rise in deceptive claims.)

The researchers explicitly instructed the AI models to stay truthful. The instruction was not followed. When the pressure to win competed with the instruction to be honest, winning won in nine of ten test cases.

The instruction to be truthful and the pressure to win are not separate systems. They share the same architecture. When they conflict, the goal wins, because the goal is what the system was built to optimize for. These models are not rogue systems or poorly built ones. They are behaving exactly as competitive optimization designed them to behave. The problem is the incentive.

That incentive comes from the same competitive dynamic that built the platforms these models run on. Social media companies discovered early in their development that content producing strong emotional reactions generated more engagement. More engagement meant more time on platform. More time meant more advertising revenue. Competitors adjusted their algorithms to match. Content that was calmer, more nuanced, or more complex generated less engagement and was gradually deprioritized. The platform that did not optimize for emotional provocation lost users to the one that did. Restraint became a competitive disadvantage.

No one at any of these companies voted to inflame the public. The competitive environment produced that result, one rational decision at a time. Researchers call this dynamic Moloch’s Bargain: when actors in a competitive system optimize........

© The Walrus