The ADD guest blog posts provide insight into the controversies and dilemmas we face in the digital age, along with ideas on how to strengthen digital democracy. Read about our relevant guest authors’ projects, activities, perspectives, experiences, ideas, and thoughts below.
Guest blogpost by Prof. Catherine Régis and Amélie Côté (1)
Without proper legal and technical oversight, AI amplifies more sophisticated and widespread information manipulation tactics and poses a risk to electoral institutions, eroding democratic processes and public trust. While electoral laws adapted to the risks of AI are needed in Canada (2) and other countries, a shared voluntary code of conduct for the use of AI by political parties offers an immediate form of protection.
Artificial intelligence as a disruptive force
Artificial intelligence (AI) has emerged as a disruptive force reshaping democracies at a global scale. Although evidence suggests it can strengthen democratic institutions — for instance, by enhancing civic engagement (3), by improving public service delivery (4), or by unmasking harmful uses of AI by political actors (5) — it nonetheless introduces significant risks to their resilience that demand urgent attention. This risk is especially patent with respect to electoral processes. From traditional resource-intensive propaganda operations to accessible and scalable manipulation tools, AI systems now enable the low-cost production of convincing deepfakes, hyper-targeted voter messaging, and large-scale disinformation campaigns. The scope is unprecedented: The International Panel on the Information Environment found Generative AI incidents in 80% of the 50 national elections held in 2024 worldwide. A quarter of the incidents were produced by identified political candidates and parties, and another half comes from untraceable sources.(6)
The Canadian Digital Media Research Network documented fabricated content mimicking trusted Canadian media sources during the last federal election in 2025, effectively distorting political discourse and confusing voters. This data revealed the country is in the early warning phase of an evolving information dynamic that could prove very damaging for public opinion, truth and electoral trust in future cycles. (7) As in other democratic countries, the spread of AI-generated political content is growing and there is limited empirical evidence on how deepfakes circulate during elections. According to a recent study, 5.86% of election-related images during the Canadian federal election were deepfakes. (8) These recent federal findings likely reflect broader patterns across Quebec and other provinces, suggesting similar vulnerabilities in Canadian electoral systems.
The Quality Over Quantity Threat of Synthetic Media
Highly realistic fabricated images pose a unique risk in the information landscape, not through sheer volume, but through their persuasive power. Despite being relatively uncommon, during the 2025 Canadian elections, these sophisticated fakes have outsized influence, particularly during politically sensitive periods. Highly credible content can shape perceptions and generate significantly more engagement than typical content, making them potent vectors for misinformation. They can easily deceive users and trigger viral sharing cascades. (9)
AI also polarizes public discourse by narrowing perspectives and reinforcing existing beliefs, while simultaneously overwhelming institutional capacity to process and respond to rapid technological changes (10). The strategic deployment of credible AI generated content within echo-chambers and close-knit communities amplifies the risk and fragments societal dialogue, as trusted networks can serve as launching pads for broader dissemination.
Current Legal Frameworks Fall Short of Adequate Response
Institutional mechanisms for attribution and correction cannot scale at the same rate as the rapid deployment of AI-generated political content. This scalability gap underscores the need for a multi-level approach that includes technological solutions, such as advanced detection systems as well as appropriate regulation – though these require time to develop and implement. Concrete solutions with immediate impact are needed, such as codes of conduct for the use of AI by political parties. Examples of these codes have already been developed in a few jurisdictions, like in the Czech Republic in 2025, in Switzerland in 2023, for the 2024 European Parliament elections, and, recently, in the Canadian province of Quebec.
While such codes cannot replace the need for technical solutions and appropriate legal frameworks, they offer the advantage of being rapid to develop and to adapt based on new AI realities and provide a much-needed complementary instrument in the AI governance toolbox. They can also cover additional ground beyond future laws by defining consensual frameworks of collaboration, training and oversight among political parties, as the Code of conduct in Quebec that we now turn to describe illustrates. Democracy
rests on an essential principle: citizens must have confidence in the authenticity of political messages and fairness of electoral rules. No voluntary code can replace comprehensive legislation. However, democracy extends beyond legal rulemaking and enforcement, and democratic resilience requires both regulatory frameworks and voluntary commitments that reflect shared values and mutual accountability among political actors.
The Quebec Code of Conduct for political parties: A Non-Partisan Democratic Safeguard
The Code of Conduct for the Use of AI by Quebec Political Parties was developed by three university professors (specialised in law, philosophy and computer science) (11) and two well-known AI organizations in the Canadian province of Quebec (IVADO and CEIMIA). It presents political parties with a unique opportunity to establish clear boundaries: AI must not become a weapon of electoral manipulation. Initially proposed to Quebec parties, this Code could subsequently be adapted for implementation across other Canadian provinces and at the municipal and federal levels.
Rather than banning AI entirely and limiting freedom of speech, the Code recognizes its democratic potential, such as analyzing citizen deliberations with safe and transparent AI systems, while preventing harmful misuse that degrades public discourse. This framework balances innovation with accountability, ensuring AI serves democratic engagement rather than undermining electoral integrity. By establishing and agreeing to voluntary standards, Quebec’s political parties can demonstrate leadership in safeguarding democratic institutions while harnessing technology’s benefits for meaningful civic engagement to strengthen electoral processes and trust.
The proposed Code outlines a comprehensive framework structured around four pillars that provide comprehensive guidance for the ethical use of AI in political contexts. The first pillar, transparency, establishes that citizens have the right to be informed whenever political content has been generated by AI systems, for example by labelling AI-generated content. The second pillar emphasizes the responsible use of AI technology, explicitly prohibiting its deployment for manipulating voters or creating fraudulent content, including fake accounts that could deceive the electorate. The third pillar addresses security and governance concerns, requiring that AI systems be adequately protected against potential data leaks and cybersecurity threats. Finally, the fourth pillar focuses on awareness and prevention, mandating that political teams receive appropriate training to recognize and avoid potential abuses of AI technology in electoral contexts.
From Research institutions to Public Awareness: The Crucial Role of the Media and Elected representatives
Traditional quality media remains a crucial pillar of our democracies while two-thirds of Canadians are concerned about AI use in electoral campaigns (with some variation across political affiliations) and continue to rely primarily on social media for information (12). The media must play a significant role together with research institutions in raising public awareness about the democratic implications of AI, especially given Meta’s permanent blocking of news access on Facebook and Instagram in 2023, which has allowed the continued consumption of information on these platforms without users’ awareness of reliability standards (13).
In Quebec’s case, the media played a pivotal role in educating the public about AI’s electoral impact and promoting the proposed Code. A journalistic coverage by one of Quebec’s leading newspapers—featuring interviews with the academic experts who developed the Code, the Chief Electoral Officer of Quebec, and political representatives—attracted widespread attention across traditional and social media platforms. This coverage pressured political parties to take official positions on the Code. Subsequently, parties that had not yet officially endorsed it faced persistent questioning from journalists at the National Assembly of Quebec regarding their hesitation. This media pressure generated significant political momentum, ultimately leading to the unanimous adoption of a motion (14) calling on all represented parties to immediately adhere to the proposed Code.
CEIMIA and IVADO are now working with the parties that supported the motion to ensure the Code fully achieves its objectives, which will soon be put to the test as the next election is scheduled for October 26th. In preparation, political parties will receive training from IVADO and CEIMIA for their staff and volunteers. The training will focus on implementing the Code and using it as a guiding framework. There will also be an ongoing monitoring of AI-related incidents that may arise during the electoral process in order to explore potential ethical, legal, and technical solutions to address these issues in collaboration with appropriate stakeholders such as political parties, civil society organizations, media outlets, and legislative authorities. The Code’s content is likely to evolve over time, ensuring it remains a relevant and effective tool for safeguarding electoral processes in the age of AI.
The challenges posed by AI to democratic institutions and processes demand immediate coordination across multiple levels of governance and society. The rapid deployment of AI-generated content significantly outpaces our current capacity for attribution and correction, creating dangerous vulnerabilities in our electoral systems that serve as fundamental pillars of democracy. The Code exemplifies the type of concrete actions we must pursue, one among many possible initiatives we can undertake in our defence of democratic values. The future of democratic systems can never be taken for granted, it must be continuously nurtured and defended through vigilant commitment and proactive measures.
References:
1. Catherine Régis is a Full professor of law at the University of Montreal, an Associate Academic Member at Mila and the director of social innovation and international policy at IVADO. She holds a Canada CIFAR Chair in AI and Human Rights as well as an International Chair in Science Diplomacy and the Global Governance of AI from the Fonds de recherche du Québec. Amélie Côté is a Knowledge Mobilization and Social Innovation Advisor at IVADO.
2. Laberge, Thomas (2026, March 8). Le DGEQ appelle à un meilleur encadrement législatif de l’IA, Le Devoir. https://www.ledevoir.com/politique/quebec/962040/dgeq-appelle-meilleur-encadrement-legislatif-ia
3. For example, between 2023 and 2025, Taiwan launched Alignment Assemblies, a nationwide deliberative initiative focused on AI governance and information integrity. Source: Fan-yu, Chen & Wang Chia-Ying (2024, October). Using AI to Strengthen Democracy: Audrey Tang on Taiwan’s Global Role. Taiwan Business Topics. https://topics.amcham.com.tw/2024/12/using-ai-to-strengthen-democracy-audrey-tang-on-taiwans-global-role/.
4. For example, Polis is an AI-powered open-source platform designed to facilitate largescale deliberative processes by clustering public opinions and identifying consensus statements. It has been used in multiple countries to inform climate policy, referendum debates, municipal decision-making and political party platforms. https://pol.is/home
5. For example, OpenFake is a large politically grounded dataset specifically crafted for benchmarking against modern generative models with high realism and designed to remain extensible through an innovative crowdsourced adversarial platform that continually integrates new hard examples. Source: Mila. https://mila.quebec/en/article/unmasking-deepfakes-with-ai
6. IPIE – International Panel on the Information Environment (2025, February). The Role of Generative AI Use in 2024 Elections Worldwide. https://cdn.prod.website-files.com/643ecb10be528d2c1da863cb/682f5ae442fffdff819ef830_TP%202025.2.pdf
7. The Media Ecosystem Observatory (2025, October). The Canadian Information Ecosystem During the 2025 Federal Election. https://www.cdmrn.ca/publications/the-canadianinformation-ecosystem-during-the-2025-federal-election
8. Livernoche, V. et al. (2026). Deepfakes in the 2025 Canadian Election: Prevalence, Partisanship, and Platform Dynamics. https://doi.org/10.48550/arXiv.2512.13915
9. Ibid.
10. Guzman Piedrahita, D. et al. (2026). AI Poses Risks to Democratic and Social Systems. https://zhijing-jin.com/d/2026-ai-risk.pdf
11. Catherine Régis, Full Professor, Faculty of Law of Université de Montréal; Marc-Antoine Dilhac, Associate Professor, Department of Philosophy of Université de Montréal; Christian Gagné, Full Professor, Department of Electrical and Computer Engineering of Université Laval.
12. The Media Ecosystem Observatory (2025, October). op. cit.
13. The Media Ecosystem Observatory (2024, August). Old News, New Reality: A Year of Meta’s News Ban in Canada. https://static1.squarespace.com/static/6450265301129e5dbabfe8a2/t/66c36dba3066f124adb4000c/1724083643261/Meta+News+Ban+Report-08-19.pdf
14. The motion is available on the Quebec National Assembly’s website (March 24th, 2026, 15:20). https://www.assnat.qc.ca/fr/travaux-parlementaires/assemblee-nationale/43-2/journal-debats/20260324/430741.html