Commons Committee Hears Chatbots Fed Quebec Voters Faulty Election Information

Canadians who turn to AI chatbots for election information cannot count on accurate answers, a House of Commons committee heard on Thursday. Aengus Bridgman, a McGill University political scientist who led a study tracking chatbot activity during the recent Quebec election, told MPs that an estimated 40 per cent of Quebec voters asked AI chatbots for information about the vote, The Canadian Press reported.

Part of the problem, Bridgman said, is that the best local sources were out of reach. “Almost all Quebec newsrooms have blocked AI chatbots from collecting their content. Which means that AI chatbots are going elsewhere for information,” he said. Taylor Owen, an associate professor at McGill’s Max Bell School of Public Policy, said researchers found that in the last three weeks of the campaign, 39 per cent of the voting advice provided by ChatGPT came from an error-filled hobby website that had itself been built with AI.

Owen stressed that nobody appeared to have acted in bad faith, whether the site’s creator, the chatbot designers or the publishers that removed themselves from AI indexes. It was, he said, a confluence of events that likely left a large number of people with false information about the election. Conservative MP Blaine Calkins called that “a little bit terrifying” and asked what could happen if someone tried it on purpose.

What it means in practice

The hearing was the first session of a new study on AI and Canada’s democratic institutions by the Commons procedure and House affairs committee. Bridgman said chatbots are effectively writing their own rules for elections, and those rules shift: during the Quebec campaign, ChatGPT went from recommending a political party one day to declining to recommend one the next. Liberal MP Braeden Caley said a chatbot had directed would-be volunteers in his August byelection to a Conservative campaign office, although the company fixed the problem once it was reported.

The witnesses offered concrete fixes. Bridgman suggested a higher standard for democratic questions, similar to the separate handling models already apply to some safety queries, since companies otherwise try to minimize the effort spent on each answer. Owen said election law itself should address what chatbots may say during campaigns. The hearing adds to concerns about AI-generated journalist personas in Canadian news, to calls for tiered AI transparency rules and to Ottawa’s recent AI transparency consultation.

In June, Parliament passed an elections bill that made it illegal to knowingly spread false or misleading information about the voting process with intent to disrupt an election, and banned deepfakes of candidates. What to watch: which AI companies the committee calls as witnesses, whether MPs recommend rules specific to chatbots during campaigns, and whether Quebec publishers revisit their decision to block AI crawlers.