Advertisement

Canadian authorities cautious, nervous about AI in political campaigns

Click to play video: 'Artificial Intelligence could threaten elections. What is Ottawa doing to regulate it?'
Artificial Intelligence could threaten elections. What is Ottawa doing to regulate it?
We’ve all heard the warnings that Artificial Intelligence (AI) could interfere with democracy. Now, 20 tech start-ups and giants including Google, Meta, TikTok, and X have signed an accord promising to work together to stop their software from creating deep fakes depicting political candidates. The companies say they’ll try to detect bogus pictures and videos but won’t ban them. And it’s still not clear how this largely symbolic pact will be enforced. David Akin looks at what one Canadian firm is doing with the technology and where Ottawa is at with regulating it. – Feb 17, 2024

As American regulators and state legislatures quickly implement new rules governing the use of artificial intelligence tools in this presidential election year, Canadian legislators and election authorities are taking a more cautious approach in a year in which three provinces will hold general elections.

And while campaign operatives in the United States increasingly explore the legitimate use of artificial intelligence to find and motivate voters more quickly and efficiently, the use of AI tools by Canadian campaigners appears to be in its infancy.

“AI is very much an emerging tool in the world of elections,” Steve Outhouse said in an e-mail. Outhouse is the campaign manager for the New Brunswick Progressive Conservatives (PCNB) and previously managed the United Conservative campaign that delivered a majority government to Danielle Smith in Alberta. He also ran both of Leslyn Lewis’s federal Conservative leadership campaigns.

In all those campaigns, Outhouse, like most modern professional campaigners, used a suite of digital tools to identify voters, raise money and get his supporters to a ballot box.

Story continues below advertisement

In New Brunswick this year, AI tools might augment existing digital tools, Outhouse said, though any phone calls the party places to voters will have a human being behind them.

“While we wouldn’t rule out potentially using elements of (AI), all calling done by the PCNB in this election cycle will either be live callers or messages recorded by the actual individual who is reaching out to voters,” he said.

A spokesperson for the New Brunswick Green Party said it, too, would stick to human beings and traditional voter contact methods.

In the United States, though, campaign operatives are already using AI tools to develop the scripts that human agents use when placing phone calls, to generate images and video, and to analyze voter data.

Click to play video: 'B.C. lawyer under fire for AI-generated fake case law'
B.C. lawyer under fire for AI-generated fake case law

One company, Votivate LLC,  is selling a tool built on artificial intelligence to non-incumbent Democrats and other “underdog” campaigns as a cheap way to match better-funded rivals. The Votivate app, the service says, gives a campaign real-time voter data, advanced modelling and analytics, and assists with door-to-door canvassing, voter calls and texting, website development and event ticketing.

Story continues below advertisement

It does not appear that anything so sophisticated is being used in the Canadian context, though some practitioners of digital campaigning are introducing AI into their workflow.

Breaking news from Canada and around the world sent to your email, as it happens.

“We primarily use artificial intelligence as an accelerator,” said Dean Tester, the president of Tester Digital, a firm that provides digital campaign services to political clients including, for example, the Saskatchewan Party.

“It helps us speed up the research process, speed up the writing process. It helps us  generate ideas.”

But Tester says finished products do not leave his shop without human oversight.

“The quality of work we get from artificial intelligence is not good enough for public consumption,” Tester said. “But it is good enough that a skilled, trained AI can take that information and use that to put together something really compelling, and in a quarter of the time that it would have taken from scratch.”

Meanwhile, Tester, Outhouse and other professional campaigners, along with federal and provincial election authorities and political parties themselves, are alert to the potential for the kind of abuse seen in the recent New Hampshire primaries when a fake robocall purporting to be from U.S. President Joe Biden was delivered to as many as 25,000 voters telling them not to vote in the primary.

The Biden voice on that robocall was created with an AI tool. Investigators have said a Texas man and his company was responsible for the calls. No link has been established between the man, his company and any registered political party.

Story continues below advertisement
Click to play video: 'Consumer Matters: Scams more sophisticated thanks to artificial intelligence'
Consumer Matters: Scams more sophisticated thanks to artificial intelligence

Even before that New Hampshire incident, legislatures in more than three dozen states had passed or are considering new laws that aim to regulate the use of AI in campaigns. The state of Oregon is the most recent. Earlier this week, the senate in that state began consideration of a bill that required political campaigns to disclose to voters when an AI tool was used to try to persuade them.

And almost immediately after that New Hampshire robocall incident, the U.S. Federal Communications Commission (FEC) prohibited the use of all  robocalls — political or otherwise — that are created with or use artificial intelligence.

And yet, the Canadian equivalent of the FEC, the Canadian Radio-television and Telecommunications Commission, believes existing rules and regulations should be sufficient to deter any such use of AI.

“While there is no specific decision or policy related to the use of AI-generated robocalls, their use could be in violation of the legislation or the CRTC’s rules, depending on the circumstances,” CRTC spokesperson Megan MacLean said in an e-mail.

Story continues below advertisement

“We would never use an AI-generated voice to impersonate a candidate, as apparently occurred in New Hampshire,” Patrick Bundrock, executive director of the Saskatchewan Party, said in an e-mail.

“We are not aware of any similar incidents happening in Saskatchewan but would be supportive of legislation or regulation to prevent this from occurring.”

Saskatchewan is also expected to see a general election this year.

Elections Saskatchewan said its chief electoral officer, Michael Boda, has been participating in discussions with other chief electoral officers around the country on AI and its potential impact on elections.

Elections Canada, for its part, is also actively analyzing whether provisions in the Canada Elections Act that relate to impersonation should be updated to better reflect the current technological landscape, a spokesperson said.

In a report it published in December, the Canadian Security Establishment warned that fake images and videos created by AI will “very likely” be used to try to undermine voters’ faith in democracy in upcoming election campaigns.

“I think everyone in politics today is concerned about how the emergence of AI and deep fakes could be used to create confusion and spread misinformation. It’s important that voters examine all information critically, and obviously campaigns need to be ready to counter false information,” Outhouse said.

Story continues below advertisement

There has only been one reported abuse of AI in a Canadian electoral context, Tester said.

In the 2023 Toronto mayoral race, a video was posted on social media that looked and sounded like candidate Brad Bradford but had him saying things he never said. It was a bit clumsy and should have been easy for many voters to spot as a fake but the tools to create those kinds of AI-powered “deep fakes” are only becoming cheaper and easier to use.

“This is the worst they’ll ever be. They’ll only get better from here,” Tester said. “So, you know, it’s it’s going to be increasingly difficult to tell the difference between what’s real and what’s not.

In B.C., where a provincial election is likely to be held in the fall, political parties there have developed a voluntary Campaign Activity Code of Practice that commits parties “to never misrepresent an artificial intelligence system as a human being.”

And whether it’s through an AI tool or more traditional means, B.C. has had a prohibition in place since last year on misrepresenting candidates and violations of that rule can result in a fine of as much as $20,000.

Dominic LeBlanc, the federal minister for democratic institutions, said his government is expected to introduce changes to the Canada Elections Act later this year that the government hopes will improve voter turnout but said he expects parliamentarians will also consider AI’s impact on elections as they consider any new legislation.

Story continues below advertisement

I worry and we have worried for a long time about the potential inappropriate use of artificial intelligence in the context of an election,” LeBlanc said. “So we would as a government be concerned about whether Elections Canada or other appropriate bodies can identify the proper way to ensure that this technology isn’t used in a way that contributes to a loss of public confidence in the electoral process.

– with files from Global’s Alex Boutillier

Sponsored content

AdChoices