Elections watchdog warned AI presents 'high' risk in current campaign: internal documents

CBC

02-04-2025

An internal briefing note prepared for Canada's election watchdog classifies the use of artificial intelligence as a "high" risk for the ongoing election campaign.
The briefing note was prepared for Commissioner of Canada Elections Caroline Simard — an independent officer of Parliament tasked with enforcing the Elections Act, including fining people for violations or laying charges for serious offences — roughly a month before the campaign kicked off.
"[The upcoming election] will quite probably generate complaints involving the use of AI tools that may constitute a contravention of the [Canada Elections Act]," the document says.
The briefing note was obtained through an access to information request by the University of Ottawa's Samuelson-Glushko Canadian Internet Policy and Public Interest Clinic and provided to CBC News.
The document — dated Feb. 23 — indicates that while AI can be used for legitimate purposes, there are risks that the tools could be used to violate election rules.
"It's important to note that the [Elections Act] does not specifically prohibit the use of artificial intelligence, bots or deepfakes. However, certain provisions under the [act] could apply if AI tools were used in ways that do contravene the [act]," a spokesperson from Simard's office told CBC News in an email.
Such violations can include the spreading of disinformation, publishing false information about the electoral process or impersonating an elections official, the spokesperson said.
Michael Litchfield, director of the AI risk and regulation lab at the University of Victoria, said there can be difficulties in going after someone who uses AI to run afoul of election rules, including finding out who they are.
"I think there's just a general difficulty with AI — and that's one of the reasons it can be misused — is identifying who is actually spreading the misinformation," he said.
The briefing note flags specific concerns about the use of AI tools and deepfakes — hyperrealistic faked video or audio.
WATCH | How AI-generated deepfakes threaten elections:
Can you spot the deepfake? How AI is threatening elections
1 year ago
Duration 7:08
AI-generated fake videos are being used for scams and internet gags, but what happens when they're created to interfere in elections? CBC's Catharine Tunney breaks down how the technology can be weaponized and looks at whether Canada is ready for a deepfake election.
"Generative AI produces convincing fakes which are rapidly debunked but can nevertheless have a significant impact," the note reads.
While the note says there has yet to be an incident where a deepfake has been used in a Canadian federal election, it points to numerous examples of deepfakes being used abroad — including one of Kamala Harris during the 2024 U.S. presidential election.
"What has occurred in elections abroad could also happen in Canada; this does not mean that it will definitely … and on a large scale," the note reads.
The document also flags that "an increase in advertising for customized deepfake service offerings on the dark web has been observed."
The impact of a deepfake can depend on how much it is circulated, the note says.
Fenwick McKelvey, an assistant professor of information and communication technology policy at Concordia University, said election rules violations are nothing new, pointing to the 2011 robocall incident.
"In situations where we had less sophisticated technology we had the same problems," he told CBC News.
But McKelvey did suggest AI adds a complicated layer to the campaign landscape.
"Generative AI arrives at a pretty dysfunctional moment in our online media ecosystem and so I don't think it's necessarily driving the challenges we face, but it doesn't help," he said.
Litchfield agreed that Elections Act violations are not new, but he said AI could exacerbate the problem.
"AI is an amplifier of these threats and makes it very easy to to create content that could run afoul of the act," Litchfield said.
One of the issues McKelvey flagged is that AI tools can be used to generate disinformation faster than it can be debunked.
"Regrettably, there's just more AI slop to replace the AI slop that we're seeing. So it's changing our media environment in ways we don't entirely know how to anticipate," he said.
During a news conference at the start of the current campaign, the head of Elections Canada raised concerns about AI being used to spread disinformation about the electoral process.
"People tend to overestimate their ability to detect … deepfakes. People seem more confident than they actually are capable of detecting it," Chief Electoral Officer Stéphane Perrault said last week.
WATCH | AI 'deepfake' election content 'a serious concern,' says Elections Canada:
AI 'deepfake' election content 'a serious concern,' says Elections Canada
9 days ago
Duration 3:13
Chief Electoral Officer Stéphane Perrault says people tend to overestimate how good they are at spotting fake and misleading election content. He says he's contacted major social media companies about the issue to ensure a 'secure election.'
Perrault also said he has reached out to social media sites such as X and TikTok to "seek their support" in combatting disinformation, specifically from generative AI.
"We'll see what action actually takes place during the election. Hopefully they won't have to intervene, but if there are issues, hopefully they will be true to their word," he said of the social media platforms.
But McKelvey is skeptical about the companies' commitments.
"Generative AI is something that platforms themselves are slightly pushing and yet we're not entirely sure how well they're actually moderated," he said.
Canada relying on 'self-regulation'
The briefing note prepared for Simard noted that Canada has generally relied on a "self-regulation" approach when it comes to AI, largely leaving it in the hands of the tech industry. But it cautions that the "effectiveness of self-regulation is contested."
"Some leading AI image generators have specific policies about election disinformation and yet, failed to prevent the creation of misleading images of voters and ballots," the document reads.
Bill C-27, which would in part regulate some of the uses of AI, was introduced in the last parliamentary session, but never made it to the legislative deadline.
Litchfield said regulations could still be passed but it will depend on the priorities of the next government. Even if something is brought forward fairly quickly, it may take some time before it's enforced.
"We are likely going to be in a regulatory vacuum for quite some time," he said. He also suggested that there could be some room to update the Elections Act itself to include AI-specific provisions.
WATCH | Debunking 3 viral election claims in 90 seconds:
Debunking election misinfo: A fake Mike Myers invoice, Poilievre's personal wealth, PPC in the polls
17 hours ago
Duration 1:21
Did Mike Myers really invoice the Liberal Party for $53,000 for his ad with Mark Carney? Is Pierre Poilievre's personal net worth really $25 million? And was a video really doctored to show the PPC in third place in the polls?
But even a regulatory framework could have its limitations, the briefing note says.
"Malicious actors seeking to sow disinformation are not likely to follow government or social media guidelines and regulations," the document says.
In a report assessing threats to Canada's democratic process released last month, the Communications Security Establishment (CSE) said known hostile actors — including China, Russia and Iran — are looking to use AI to fuel disinformation campaigns or launch hacking operations.
It said these actors "are most likely to use generative Al as a means of creating and spreading disinformation, designed to sow division among Canadians and push narratives conducive to the interests of foreign states," wrote the agency in its report.
"Canadian politicians and political parties are at heightened risk of being targeted by cyber threat actors, particularly through phishing attempts."
Concerns that legitimate use of AI could spark complaints
There are already examples of AI being used to spread misinformation in this campaign.
An obscure website featuring articles that appear to be AI-generated has been pumping out dubious information about party leaders' personal finances. There have also been fake election news ads attempting to lure Canadians into sketchy investment schemes. Some of those ads have been taken down.
McKelvey said the use of AI is also leading to an increase in "news avoidance."
"We're now feeling less and less trust in any content we see online, whether it's AI-generated or not. And that's something that's going to make it harder for credible information sources to be believable," he said.
WATCH | How we busted this fake election content luring you into get-rich-quick schemes:
How we busted this fake election content luring you into get-rich-quick schemes
5 days ago
Duration 3:07
A network of shady websites is using ads promoting fake news articles focused on election issues to lure people in. The CBC News visual investigations team looked into it — and got some taken down.
McKelvey's concern is echoed in the briefing note prepared for the commissioner.
"[Deepfakes] contribute to affecting the public sphere by confusing people about what is real and what is not," it reads.
The briefing note also warns the commissioner that the use of AI is likely to spark a number of complaints during this election campaign, even in instances where no rules have been broken.
"The resulting cases could be complicated to assess and may be on a large scale," the note reads.
But McKelvey said the use of AI for benign purposes can alter the ways in which campaigns are conducted. He pointed to U.S. President Donald Trump posting an AI-generated image to social media that depicted him standing next to a Canadian flag overlooking a mountain range as an example of something that doesn't break any rules, but is "weird" nonetheless.
"There's something stranger here when it comes to AI-generated content where it kind of allows for the expression of strange political ideas or the kind of normalization of this untrue, unreal content," he said.