Many Quebec voters got false election information from chatbot, MPs hear

OTTAWA — Canadians turning to chatbots for election information don’t always get accurate responses, a parliamentary committee heard Thursday.

An estimated 40 per cent of Quebec voters asked AI chatbots for election information, said Aengus Bridgman, a McGill political scientist who led a study tracking chatbot activity during the recent Quebec election.

But because most Quebec news outlets don’t allow chatbots to access their content, he said, those voters couldn’t obtain the most accurate and up-to-date information.

“Almost all Quebec newsrooms have blocked AI chatbots from collecting their content. Which means that AI chatbots are going elsewhere for information,” he said.

News outlets in both Canada and the United States have opposed the use of their content by AI platforms without compensation and permission.

Taylor Owen, associate professor in the Max Bell School of Public Policy at McGill, said researchers found that in the last three weeks of the election campaign, 39 per cent of voting advice provided by ChatGPT came from an error-filled hobby website built by AI.

“There was nothing nefarious by this person’s actions nor necessarily by the design of the chatbots themselves, nor by the publishers who de-indexed themselves,” he said.

“It was just this confluence of events that led to, likely, a large number of people getting false information about the election.”

“That’s actually a little bit terrifying … If that can happen by accident, imagine what could happen potentially on purpose,” Conservative MP Blaine Calkins responded.

Owen said that’s the question that needs to be asked “quite urgently.”

Thursday’s meeting was the first for a new study on AI and Canada’s democratic institutions launched by the House of Commons procedure and House affairs committee.

Bridgeman told the meeting chatbots are writing their own rules on what they can do during an election. He noted those rules can change — for instance, during the Quebec campaign, ChatGPT would flip-flop its responses when asked to recommend a political party.

“It went from one day to the next, recommending a political party to not recommending a political party,” he said.

Liberal MP Braeden Caley, who was elected in August, said his campaign ran into an issue during his byelection where a chatbot was directing potential volunteers to a Conservative campaign office.

He said the company behind the unnamed chatbot was “actually very helpful in solving this once it was reported.”

Bridgeman said one way to address the issue of chatbots giving false information would be to require higher standards for election-related questions, like the standards companies must meet for queries involving safety issues.

“There’s a degree of effort that chatbot will assign to each query, and that effort is, in the interest of the company, attempted to be minimized. So they are going to seek the least information possible to still provide what is a satisfactory, in their view, answer to the question,” he said.

“A very concrete thing to do is to say, look, for democratic questions, in the same way that for certain safety and other issues, there is a different path that model will go through.”

Taylor said elections law should address how chatbots provide election information.

“We accept very strict limits on speech, during this limited time period of democratic decision-making and deliberation, and I think chatbots are going to have to be included in that,” he said.

“What kind of speech from AI systems are we going to allow during our elections? That is a thing that can be legislated.”

In June, the federal government passed an elections bill that made it illegal to knowingly spread false or misleading information about election activities or the voting process with the intent to disrupt an election or affect the outcome. It also banned digital deepfakes of candidates.

This report by The Canadian Press was first published Oct. 8, 2026.

Anja Karadeglija, The Canadian Press