Risques de l'IA pour l'intégrité électorale et informationnelle au Canada

Generative AI is creating concrete threats to the integrity of Canadian elections at both the federal and provincial levels. During the 2025 federal election, AI-generated deepfake videos of Prime Minister Mark Carney reached millions of viewers on TikTok, Facebook, and X. Over 40 Facebook pages ran fraudulent investment scams using AI-generated likenesses of Carney and Dragon's Den personalities. Academic analysis documented the prevalence and platform dynamics of election deepfakes.

Canada's intelligence agencies have assessed the threat as significant and growing. The Communications Security Establishment's 2023 update on cyber threats to Canada's democratic process identified generative AI as making it easier for state and non-state actors to produce convincing disinformation. The Hogue Commission's final report on foreign interference identified AI-enabled disinformation as part of the broader threat landscape. CSE noted that the barrier to creating high-quality synthetic content has dropped substantially.

The legislative and institutional response has not kept pace. The Canada Elections Act was drafted before generative AI existed. While it prohibits certain misleading communications, it does not address synthetic media. The Chief Electoral Officer proposed targeted amendments in November 2024 but no legislation has been introduced. Elections Canada lacks dedicated technical capacity for synthetic media detection.

At the provincial level, Quebec's Chief Electoral Officer (DGEQ) has publicly identified AI as a serious threat to the October 2026 provincial election while acknowledging his institution's limited capacity to respond. Bill 98, adopted in May 2025, created an offense for knowingly spreading false election information with penalties up to $60,000 — but the DGEQ concedes that prosecution under the criminal standard of proof is extremely difficult. Élections Québec received complaints from citizens who obtained incorrect election information from commercial AI chatbots during municipal elections.

The Commission de l'éthique en science et en technologie (CEST) has documented that AI-generated deepfakes disproportionately target women through non-consensual pornographic content, potentially discouraging their political participation — adding a gendered dimension to the election integrity hazard.

The pattern is consistent across jurisdictions: institutional threat assessments identify AI disinformation as significant, but the governance response — legislative frameworks, detection capacity, platform obligations — lags behind the capability that enables the threat.

Major platforms have implemented election integrity policies, including labeling requirements for AI-generated content, restrictions on political advertising, and partnerships with fact-checking organizations. Some AI-generated deepfakes during the 2025 election were identified and labeled by platforms and journalists relatively quickly. The debate centers on whether voluntary platform measures and existing election law provide adequate protection, or whether AI-specific electoral provisions are needed.