Authors:

Aqsa Farooq
Claes de Vreese
Co-chairs of the EDMO Internal Expert Group on Generative AI and Disinformation

The 2024 global election cycle unfolded against the backdrop of rapid advancements in generative AI. While these technologies hold immense potential for enhancing communication and accessibility, they also pose new challenges to the integrity of democratic processes. This was of particular importance in a year that saw almost 2 billion people from over 70 countries eligible to vote in elections, including the 2024 European Parliamentary elections. As a result, the potential threat of AI-generated disinformation proliferating voters’ online information spaces and social media feeds was a frequent point of discussion.

This topic was also the focus of EDMO’s Internal Expert Group on Generative AI and Disinformation this year. Co-chaired by Claes de Vreese (University of Amsterdam) and Aqsa Farooq (University of Amsterdam), this group met throughout the year to discuss key challenges, opportunities and trends in relation to the intersection of Generative AI and disinformation from multiple perspectives. For its final meeting of the year, it was only fitting to discuss how the unfolding events of the global election year shaped the conversation around AI-generated disinformation, focusing on the following questions: How much of an impact did generative AI have on elections this year? Are we overestimating its influence, or is it a genuine threat to democracy?

For this discussion, the chairs of the Expert Group partnered with the AI, Media and Democracy lab to organise an online session, featuring speakers with expertise on the topic. The session was held on November 25 and the speaker panel included Mark Scott (Atlantic Council), EDMO’s own Mato Brautović (ADMO, University of Dubrovnik), and Taberez A. Neyazi (National University of Singapore).

Mark Scott shared a U.S.-centered perspective on the evolving role of AI in elections. Contrary to sensationalist fears of deepfakes dominating electoral discourse, Scott highlighted the growing use of large language models (LLMs) to analyse voter data and target campaign messaging. These tools have proven highly effective for reaching specific audiences and are poised to become even more significant in future campaigns.

Scott also noted examples where AI was harnessed positively. In Belarus, a chatbot provided voters with information about opposition parties without risking activists’ safety—a testament to AI’s potential for empowering democratic movements. However, he cautioned that foreign actors have also weaponised generative AI to interfere in elections, underlining the need for both vigilance and innovation.

Turning to Asia, Taberez A. Neyazi described how AI has shaped electoral campaigns in India and Indonesia. In India, deepfakes surfaced in controversial ways, such as fabricating statements by politicians. While the impact of such disinformation on voting remains uncertain, it highlights the dual-edged nature of generative AI. On the brighter side, voice cloning technologies allowed politicians to address voters in their native languages, bridging communication gaps and fostering inclusivity in a linguistically diverse nation. Neyazi emphasised that AI has immense potential to make democracy more accessible, provided it is used responsibly.

Mato Brautović offered a European perspective, focusing on Croatia’s parliamentary elections. AI-generated content, including deepfakes, targeted prominent politicians. While the quality of such disinformation was generally poor and recognisable by informed users, the lack of labeling by social media platforms and insufficient action from fact-checkers was concerning. Brautović pointed out that the rapid spread of disinformation facilitated by AI creates an urgent need for stronger detection and regulation mechanisms. ADMO hub’s report on the use of Generative AI in relation to the Croatian elections can be accessed here.

In the panel discussion, co-host Sophie Morosoli (AI, Media and Democracy lab) raised a pivotal question: Is Generative AI simply accelerating pre-existing threats to democracy, or does it represent a novel challenge?

The consensus among the speakers was that AI acts as an amplifier of existing issues, such as polarisation and declining trust in democratic institutions. Brautović noted that the speed and linguistic accessibility enabled by AI make it a powerful tool for spreading disinformation, particularly when used by foreign actors. Scott stressed the importance of demystifying AI to curb unnecessary fears while addressing legitimate risks.

Looking forward, the panelists suggested several initiatives to mitigate these risks:

Watermarking AI-generated content: Scott viewed this as a temporary measure but emphasized the need for long-term solutions like fostering digital literacy.

Transparency and accountability: Brautović argued for stronger obligations on developers to consider societal impacts during the creation of AI technologies.

Global knowledge sharing: Neyazi highlighted the value of learning from international experiences, suggesting that the challenges faced in one region could provide critical lessons for others.

While the misuse of generative AI in disinformation campaigns poses a serious threat, this technology is not inherently harmful. The positive examples of chatbots in Belarus and voice cloning in India demonstrate AI’s potential to strengthen democratic engagement. However, realising this potential requires proactive measures: robust regulations, enhanced digital literacy, and global collaboration.

As we look ahead, generative AI is likely to remain a double-edged sword. By fostering transparency and accountability, while leveraging AI’s positive applications, democracies can adapt to this evolving landscape. The lessons from 2024 make one thing clear: AI is not just shaping the future of elections—it is shaping the future of democracy itself.

We would like to thank all of the speakers, as well as the AI, Media and Democracy lab for their contribution to the success of the event. We look forward to taking these conversations further into 2025.