Stopping AI ‘slop shots’ in modern politics

Hockey is Canada’s favourite sport. Any wonder then that our political parties have decided it’s time to take AI “slop shots” at their opponent’s net?

The Conservative Party of Canada was first and strongest out of the dressing room. The other week, it released a fully AI-generated video ad about food banks and hungry people lining up with empty shopping carts. Like much political advertising, this falls in the category of being accurate without being true.

Food Banks Canada statistics say there were 2.1 million monthly visits to food banks in the country in March 2025. Food bank usage rates have doubled, they say, since 2019. So, the issue exists but the talking “Canadians” in the ad do not. They are AI-generated avatars, not real people.

Welcome to the new frontier of campaign advertising. It is real even if the content it produces is not. If you’ve ever watched American TV during an election year — or just the drama series West Wing — you saw a version of this after every ad spoken by a candidate: “I approve this message.”

That requirement comes from the Stand By Your Ad provisions of the 2002 Bipartisan Campaign Reform Act, itself a response to increasingly negative attack ads financed by candidates, parties and third-party funded interest groups and so-called SuperPACs. The legislation was meant to take the temperature down in political campaigns since throwing mud at your opponent always leaves some on you, too.

For years now, election materials in Canada have required a formal authorization statement. This is meant not to clean up the message but to clean up the money. It ensures campaigns stay within their legal spending limits. Provisions such as this in our election administration laws are meant to make campaigns fair and transparent.

AI advertising changes all this. It’s cheap and increasingly easy to manufacture, so spending limits are no barrier. That makes it a campaign manager’s dream but potentially a voter’s nightmare. Plus, AI is already demonstrating an uncanny ability to create easily digestible messaging that is far closer to voter mindsets than the traditional route of polling, focus group testing and ad-agency-created videos. This makes them that much more impactful and influential in driving voting behaviour.

But they also make a mockery of fair and transparent election messaging. No matter that an AI ad may be “authorized” by an official campaign. The images, voices and words are not real. They are, in the vernacular, “deepfakes.” In the hands of bad actors, foreign and domestic, they sow dissent and disruption about democracy, candidates and governments. As Mark Twain said, “A lie can travel halfway around the world and back again before the truth can put on its boots.”

How is this different from campaigns today, you may ask?

Election ads all stem from a form of creative licence and, let’s say, flexibility in context and facts. Microtargeting of voters via data mining and social media algorithms exists now. AI simply shrinks the time, space and cost needed to produce a traditional political ad viewed on television or social media.

AI campaign advocates consider opponents to such ads modern-day Luddites. This is the same type of handwringing and tut-tutting, they contend, that accompanied the advent of radio or television or the Internet.

This is a cavalier, slippery-slope attitude. The impact that matters most is on voters, not campaigns. Making campaigns easier and more effective is irrelevant to the resultant impact on democratic trust and confidence in election outcomes.

The issue is not just AI-generated ads. It’s also about AI engagement with voters. A study published in 2025 of large language models as a persuasion tool in last year’s federal election in Canada found a 10-point increase in candidate preferences. AI was far more effective in shifting voter attitudes towards a candidate than traditional human persuasion. Put another way, a machine is becoming better at convincing a human to vote for another human than any human could.

None of this means elections as we know them are relics of the past. But the accepted determination of what constitutes legitimate and ethical election discourse has already arrived.

Manitoba’s decision to amend its election law to ban deepfakes — AI-altered images and audio recordings of candidates — is a strong step to regulate this corruption and protect voters. So far, no one else has followed suit.

Around the world, though, in 11 jurisdictions, including the European Union, we are seeing the emergence of codes of conduct to voluntarily govern the use of AI in elections.

This would be a good place to start in Canada. Political parties are notoriously reluctant to accept government regulation and proscriptions on their behaviour. The next best thing would be for them to step up and set up a common code of conduct for how they will use AI in campaigning.

That code would have three main components: a prohibition on using deepfakes, public disclaimers on AI-generated material and guarantees of data security by not using foreign brokers.

The AI revolution has arrived. Our political parties need to step up to stop the AI slop shots in politics. It’s their obligation to us, the voters.

David McLaughlin is a former clerk of the executive council and cabinet secretary in the Manitoba government.