As generative artificial intelligent (AI) systems become widely available and accessible, public concern about the impact of AI on democratic societies rises, particularly regarding their potential to influence voters’ attitudes and behavior.
It is not a hypothetical risk, as AI-generated content is already influencing electoral processes and narratives around the world: in the United States, robocalls impersonating President Joe Biden’s voice discouraged people from voting in the New Hampshire primaries in 2024; in Slovakia, a viral fake audio that circulated on Facebook showed the leader of the liberal Progressive Slovakia party discussing how to rig the parliamentary election; during the European Parliament elections in June 2024, a deepfake video in Germany showed Chancellor Olaf Scholz calling for a ban on the Alternative für Deutschland (AfD) party; in Bangladesh, a fake video announced a political candidate’s withdrawal from elections; in South Africa, a deepfake video showed U.S. rapper Eminem endorsing the Economic Freedom Fighters party; in Argentina, candidates Javier Milei and Sergio Massa widely shared AI-generated images and videos in their campaigns in 2023; in Romania, probably the most notable example, the first round of the 2024 elections was annulled by the Romanian Constitutional Court, in what was considered by the New York Times “the first major election in which A.I. played a decisive role in the outcome”.
In response to this context, Brazil has introduced new restrictions for the use of AI during the electoral period for the upcoming general elections, to be held in October 2026. In early March 2026, the Brazilian Superior Electoral Court (TSE) published Resolution n° 23.755, updating the country’s electoral advertising rules. The resolution builds on the framework established for previous elections and introduces several notable changes aimed at addressing the electoral harms posed by AI and generative conversational AI systems, such as ChatGPT (Open AI), Claude (Anthropic), Copilot (Microsoft), and Gemini (Google).
This post examines the Brazilian approach as an example of sectoral AI regulation in elections, discussing how it compares to other jurisdictions and what it reveals about the possibilities and limits of context-specific responses.
Why Did Brazil’s Electoral Court Step In?
The Brazilian electoral governance system relies on the role of tribunals in its supervision and the TSE is the highest structure within the Electoral Justice system. According to the 1988 Constitution and the Electoral Code, the Court encompasses jurisdictional, administrative and normative or regulatory powers. Through its normative activities, TSE issues binding instructions for the enforcement of electoral rules – which is the legal basis for the resolutions discussed here.
As digital technologies have become increasingly embedded in electoral dynamics, the Court has been called upon to respond to new forms of interference, manipulation, and disinformation that threaten electoral fairness. These challenges came to the forefront during the 2018 presidential election, widely regarded as a disruptive moment in Brazilian democracy. For the first time in the country’s democratic history, a candidate who did not meet the traditional prerequisites for electoral success (substantial campaign funding, airtime in electoral broadcasting, and a network of political support across states) was elected, amid a campaign marked by an unprecedented volume of digital disinformation. The 2022 general elections were also marked by the rise of content criticizing and attacking democratic and constitutional institutions such as the TSE, the Federal Supreme Court (STF), and the electronic ballot box, following a pattern of constitutional erosion that marked Jair Bolsonaro’s government.
This gradual decline in the quality of Brazilian constitutional democracy reached a breaking point on January 8, 2023, when Bolsonaro supporters stormed the buildings of all three branches of power in Brasília, claiming the electoral process had been rigged. The attacks were not an isolated episode. They were the most visible expression of a longer process of institutional delegitimization, fuelled by disinformation and sustained attacks on the credibility of the electoral system, in which digital platforms played an important role. Generative AI now adds a new dimension to this landscape, as it amplifies the capacity to produce and spread convincing false content at scale, but, more importantly, it provides a powerful persuasion tool.
From Deepfake Bans to Chatbot Restrictions: The TSE’s Regulatory Evolution
Since the 2018 election, the TSE has launched a series of institutional initiatives to address these emerging threats for elections. For example, it has implemented the Electoral Justice Permanent Program on Countering Disinformation, aiming to reduce the harmful effects of disinformation. Through its normative powers, TSE issued resolutions to address disinformation related to the integrity of the electoral process (Resolution n° 23.714/2022), as well as to regulate the use of artificial intelligence in electoral campaigns and political communication on social media platforms (Resolution n° 23.732/2024). The 2024 Resolution established, among other measures, the absolute prohibition of deepfake use in electoral advertising, the requirement of identification labels for AI generated content and the restriction of the use of chatbots and avatars to mediate campaign communication.
The rise of generative AI systems has presented a new and pressing challenge for elections. Recent studies suggest that conversations with AI are more effective at shifting people’s political views than traditional political advertising, a finding that underscores the urgency of the regulatory response. Therefore, for the 2026 general elections, the TSE introduced norms aimed directly at these harms. Resolution n° 23.755/2026 builds on the 2024 framework and introduces, among other things, a notable change: the prohibition of AI application providers, such as chatbots and AI assistants, from ranking, recommending, or expressing any form of electoral preference, even when directly asked by the user.
This is a meaningful step. As AI tools become increasingly integrated into how people search for and consume information, the risk of these systems influencing political opinion – and, consequently, acting as relevant political actors – becomes very real. The concern is not only about disinformation, but about the structural influence that automated, personalised, and opaque responses can have on voters at scale.
Importantly, the TSE’s approach reflects the logic of sectoral regulation: tailoring legal obligations to the specific risks and characteristics of AI applications within particular domains. This is especially important in high-risk contexts, where the misuse of AI can have immediate and far-reaching consequences for public trust and democratic stability. By focusing not on the technology itself, but rather on the context of use, the application and the potential for harm, this approach enables more proportionate, effective, and context-sensitive responses.
AI and Elections Across Jurisdictions: Comparative Lessons and Challenges
Brazil is not alone in grappling with these challenges, but its approach stands out in comparative perspective. Several jurisdictions have introduced measures targeting AI-generated content in elections.
South Korea, for example, amended its Public Official Election Act in 2024 to ban realistic AI-generated media in the 90 days before the election. In Singapore, the Elections (Integrity of Online Advertising) Bill was amended to prohibit manipulated online election advertising containing realistic but false representation of a candidate. Some U.S. state legislatures have adopted a patchwork of disclaimer and/or timing requirements and prohibitions. These measures, however, tend to focus narrowly on deepfakes and manipulated content.
In the EU, the AI Act, despite its broader normative ambition, does not explicitly address electoral applications of AI or establish mitigation measures tailored to electoral risks. On the other hand, the Digital Services Act (DSA) and the EU Regulation on the Transparency and Targeting of Political Advertising (TTPA) address systemic risks for electoral processes but are not AI focused.
Brazil’s framework, by contrast, is explicitly calibrated to the dynamics of the electoral process and to the high risks posed by AI applications in this important domain. As a sectorial strategy, it also reduces the risks of regulatory overlap or unnecessary burdens.
That said, the Brazilian model has real limitations. Operating without a general AI law, the TSE’s resolutions function in a fragmented regulatory environment, one that lacks the structural safeguards needed to address systemic risks such as political microtargeting or opaque recommendation systems. The effectiveness of the framework also depends heavily on judicial interpretation, platform cooperation, and institutional capacity.
Evidence from Brazil’s municipal elections in October 2024 illustrates these difficulties: despite the existing ban on deepfakes, AI-manipulated content circulated widely across social media platforms. This included deepfake pornography targeting female candidates, impersonations of well-known journalists, and fabricated videos designed to mislead voters about electoral information, showing that the gap between regulatory ambition and enforcement remains significant.
Conclusion
Brazil’s approach offers a concrete, though still imperfect, response to the intersection of AI and electoral law. By tailoring legal obligations to the specific risks of AI in elections, the TSE’s framework goes further than most international experiences, which tend to be either too broad or narrowly focused on deepfakes. It combines restrictive measures with proactive obligations, reflecting a logic of sectoral regulation that is well-suited to high-risk domains.
Yet sectoral regulation alone is not enough. Without a general AI law, Brazil’s framework operates in a fragmented legal environment, leaving critical issues such as algorithmic transparency, auditability, and political microtargeting without adequate legal grounding. Its effectiveness also depends on judicial capacity, platform cooperation, and institutional coordination, factors that remain uncertain ahead of the 2026 elections.
Brazil’s experience can serve as both a reference and a warning: while context-specific regulation may provide timely safeguards, it cannot replace the need for comprehensive, systemic approaches to AI governance rooted in democratic values and constitutional principles. The most promising path likely lies in combining both: pairing overarching principles that guide the broad use of AI with sectoral norms that address the concrete risks of specific applications.
Front Image: Gemini generated illustration
Image 1: Photo by Roque de Sá/Agência Senado. Source: Wikimedia Commons. License: CC BY 2.0.
Image 2: Photo by Alejandro Zambrana/Secom/TSE. Source: Wikimedia Commons. License: CC BY 2.0.
Suggested citation:
Ana Luiza Marques, ‘AI and Elections: Brazil’s Sectoral Regulation in Comparative Perspective’ (Comparative Digital Law Blog, 02 June 2026) <https://lawandtech.ie/ai-and-elections-brazils-sectoral-regulation-in-comparative-perspective>.
About the author:
Ana Luiza Marques is a PhD researcher in Law at the Federal University of Minas Gerais (UFMG), Brazil, and a visiting doctoral researcher at Dublin City University (DCU), Ireland. Her research focuses on judicial content moderation and digital constitutionalism. She also works as a litigation lawyer in Brazil. Her broader research interests include democratic erosion, platform regulation, and the intersection of technology and democratic institutions.
