AI Is Changing Democracy. The Question Is How We Respond.

Beyond classrooms and the workplace, artificial intelligence is quietly infiltrating a more precarious corner of society: the information environment that democracies depend on. As generative AI becomes increasingly accessible, questions about electoral integrity, public trust, and democratic accountability are becoming impossible to separate from conversations about the technology itself.

At a recent Mila/Techaide event in Montréal, Prof. Catherine Régis explored this evolving relationship between AI and democracy. Rather than framing AI as an existential threat, she argued that it acts as an accelerator. Problems like misinformation and declining institutional trust are not products of AI, but the technology makes these issues more difficult to manage. As such, Régis focused on elucidating what governments, political institutions, and citizens can do to respond to these risks before they become entrenched.

AI Amplifies Existing Democratic Challenges

The discussion began with a broader tension facing democracies today. Although democracy remains the form of government people trust most, confidence in democratic institutions has been gradually declining. Concerns about political polarization and the growing influence of large technology companies have all contributed to this current climate of erosion.

Within this context, generative AI raises the stakes because it dramatically lowers the cost of creating convincing content. Text, images, audio, and video that once required significant time or expertise can now be produced in seconds by virtually anyone. The sheer volume and improved quality of AI-generated content is making it increasingly difficult for people to know what is true, and what deserves their trust.

In fact, Canada’s Foreign Interference Commission has already identified AI-enabled disinformation as an emerging risk to democratic institutions. As these tools continue to improve, governments are being forced to consider how existing safeguards hold up in an environment where manipulation becomes easier and faster.

Elections Offer a Preview Of What is at Stake

Examples from recent elections around the world illustrate how AI is already influencing democracy.

In the United States, AI-generated robocalls imitating Joe Biden’s voice attempted to discourage voters from participating in the Democratic primary. During Mexico’s presidential election, fabricated videos circulated online containing false voting instructions. Romania experienced cyberattacks targeting electoral authorities alongside broader interference campaigns that contributed to the annulment of the first round of its presidential election.

Although the examples themselves are different, they reveal the same pattern: AI allows deceptive content to be created quickly, distributed widely, and tailored to specific audiences. As a result, even after misinformation is disproven, its effects often remain pervasive. Public uncertainty does not disappear simply because the facts are eventually corrected.

Beyond the creation of misleading content itself, identifying who is responsible has become difficult as well. Various parties including foreign governments, political organizations, anonymous online actors, and coordinated networks can all contribute to the spread of misleading content, which often leaves institutions with few clear avenues for accountability.

Democracies Depend On Trust

Another critical concern lies in what happens when the uncertainty created by AI fabricated content becomes widespread.

Democratic systems depend on citizens believing that elections are legitimate, and that reliable information remains accessible. AI threatens that confidence not only because it can generate convincing falsehoods, but because it encourages people to question authentic information as well. This ubiquitous uncertainty means that distinguishing fact from fiction is only part of the problem. More significantly, confidence in democratic institutions begins to weaken alongside it. From this perspective, the greatest risk is the gradual erosion of the trust that allows democratic systems to function.

Responding Before the Problem Grows

A recurring theme throughout the discussion was the need to prepare before AI-enabled threats become more difficult to contain. Democratic institutions need proactive mechanisms that allow them to respond as AI continues to evolve.

Régis discussed an international policy statement developed by researchers and experts from several countries that proposes practical responses to AI in elections. Some recommendations focus on updating legislation to address deceptive uses of generative AI during campaigns, while others recognize that regulation alone cannot keep pace with technological change. Political parties also need shared expectations around how AI should be used during elections. When common standards are established, it becomes much easier to distinguish legitimate innovation from practices that undermine public confidence.

She also highlighted the importance of specialized teams capable of responding quickly to AI-related incidents. Because election interference unfolds so quickly, governments often have little time to rely on lengthy institutional processes. Developing dedicated expertise would improve their ability to detect emerging threats, investigate them efficiently, and coordinate effective responses before they gain momentum. Concurrently, the cross-border nature of many AI-enabled campaigns means that these efforts will ultimately also depend on stronger international cooperation rather than the actions of any one democracy alone.

Quebec’s Voluntary Approach

Quebec provides one example of how institutions can respond proactively.

Recognizing that legislation often moves more slowly than technology, Régis worked alongside political parties to develop a voluntary code of conduct governing AI use during election campaigns. Rather than imposing rules externally, the process invited parties themselves to establish shared expectations.

The code promotes transparency around AI-generated political content while discouraging deceptive practices intended to mislead voters. It also creates an ongoing governance structure that allows participating parties to revisit the guidelines as the technology changes.

The value of this approach is that it reduces incentives for escalation. Without shared standards, increasingly deceptive AI tactics adopted by one political party could quickly pressure its competitors to follow suit simply to remain competitive, creating a cycle that becomes progressively more difficult to reverse. By establishing common commitments early, the code helps prevent those dynamics from becoming normalized.

She also pointed to an unexpected benefit: public discussion of the code prompted journalists to ask political leaders whether they supported it, extending accountability beyond legislation through sustained media scrutiny.

AI Literacy Is Becoming A Democratic Skill

The final part of the discussion shifted attention from institutional responses to the role of citizens themselves.

As AI-generated content becomes increasingly convincing, technical safeguards alone will no longer be enough to protect democratic discourse. People also need the ability to evaluate and discern information critically, understand how AI-generated material is created, and recognize when online content deserves greater scrutiny.

In Régis’ view, Canada still has considerable ground to cover in this area. Public understanding of AI has not kept pace with the technology’s rapid adoption, even as other jurisdictions—particularly in Europe—have begun treating AI literacy as a core element of democratic resilience alongside regulation itself.

Although improving public understanding will not eliminate misinformation, it can leave societies far better equipped to recognize and respond to misleading content when it inevitably emerges.

Looking Ahead

AI has moved beyond the boundaries of a purely technological debate, raising broader questions about how democratic institutions should adapt to a rapidly changing information environment.

The technology itself does not carry a predetermined impact. Its consequences will be shaped by the choices made by those who develop, regulate, and engage with it.

Ultimately, the vulnerabilities facing democracies are significant, but they are not unavoidable. As Régis argued throughout her talk, protecting democratic institutions will require a proactive approach that goes beyond reacting to individual AI-generated controversies and focuses on creating the capacity to adapt as the technology continues to evolve.

Acknowledgement

The ideas discussed in this article were informed by Prof. Catherine Régis’ presentation at Mila/Techaide, which examined the evolving relationship between artificial intelligence and democracy. Her work explores questions around responsible AI governance, policy development, and the societal implications of emerging technologies. More information about Prof. Régis’ research can be found through her profile at Mila: Catherine Régis | Mila.

Related Links

https://apnews.com/article/new-hampshire-primary-biden-ai-deepfake-robocall-f3469ceb6dd613079092287994663db5

https://www.bbc.com/news/articles/cn4x2epppego

https://www.freiheit.org/publikation/generative-ai-and-its-influence-mexicos-2024-elections-0