Electoral Deepfakes and Democratic Integrity in Argentina: Reflections on Building an AI Accountability Agenda

A project reflection by Dr. Damián Loreti and Dr. Luis Lozano, supported by the Pulitzer Center through the South to South AI Accountability CoLab microgrant. 

In the six months after the 2025 elections in Argentina, two researchers invited 40 members of academia, civil society, and journalism to participate in a community of practice. This group would debate the effect of deepfakes produced with AI on politicians during the electoral processes. 

In Argentina, deepfakes in electoral contexts are situated at the intersection of AI governance, digital platform regulation, and democratic integrity in the region. This topic connects human rights standards from the Inter-American system with specific electoral case law and institutional practices, providing a practical perspective on the impact of AI-generated manipulation on campaigns and voter information. 

We produced a paper from the debates that systematically reconstructs cases from 2005 to 2025, focusing on recent deepfake incidents involving key figures in the Argentinian national and local political landscape that were produced with AI. The project demonstrates how AI-enabled disinformation has evolved from sporadic occurrences to a pervasive threat to democracy.

At the same time, our paper situates Argentina’s response within broader debates on fake news, election misinformation, and platform responsibility. In the Argentinian context, AI accountability cannot be reduced to technical fixes but must be framed through freedom of expression and democratic safeguards. The paper denotes several recommendations through policy briefs to inform key stakeholders.

Why the project matters for accountability

The project is relevant to holding powerful actors accountable because it identifies the specific ways in which political actors, parties, consultants, and platforms exploit generative AI to manipulate voters under the cover of plausible deniability, regulatory gaps, and permissive judicial practices. By showing how deepfakes and cheap fakes are strategically released in “last-minute” windows like campaign closures and electoral veda, it exposes how informational asymmetries are used to game the system while institutions struggle to respond in time.

It also clarifies the responsibilities of different actors: courts and electoral authorities using article 140 of the National Electoral Code, prosecutors coordinating with cybercrime units, platforms claiming neutrality while exercising de facto editorial power, and political leaders amplifying misleading content. In doing so, the research supports a rights-based accountability agenda that resists both state censorship and platform self-regulation without oversight, instead pushing for enforceable duties of transparency, cooperation, and preservation of evidence in line with Inter-American standards. 

The community of practice united several groups; addressing deepfakes is a matter that involves politicians, judicial authorities, journalists, and requires proper systematized knowledge. A common space has shown the limits of the knowledge, time, and resources each sector holds, and the urgency to work together and join forces in order to develop useful tools that will allow the public to trust electoral information during campaigns. 

The process allowed us to go beyond case-by-case commentary and produce a comprehensive mapping of deepfake-related incidents, jurisprudence, and regulatory initiatives across a full electoral cycle, something that is rarely feasible within routine academic and teaching workloads.

Cross-disciplinary accountability work and its network of journalists, technologists, and researchers helped us connect Argentine developments with international tools and experiences, such as the Election Deepfakes Dataset, Arbiter, SMAT, DisTrack, and simulation projects like SimPPL. These references informed our recommendations for a national toolkit of early warning and narrative tracing capacities embedded in the electoral justice system, rather than outsourced to private platforms or ad hoc initiatives. Finally, the network opened avenues for future collaboration with fact-checkers, media organizations, and other Latin American teams, allowing the project’s outputs to circulate beyond academic circles and influence policy conversations in a more timely way. 

Challenges: Political climate, AI hype, and how we coped

The implementation of this project unfolded within a challenging political climate in Argentina, marked by polarization, economic crisis, and a government that simultaneously invokes freedom of expression and explores more punitive approaches to speech and media regulation. The global hype around AI—often framed either as a panacea or an existential threat—complicated efforts to discuss deepfakes in a grounded way, since public debates tended to oscillate between alarmism and denial rather than focusing on concrete institutional reforms. 

This context made our work harder: Access to official information was uneven; institutional interlocutors were often overwhelmed by immediate crises; and any proposal touching electoral content risked being read through partisan lenses. We tried to overcome these obstacles by anchoring the analysis in Inter-American human rights standards, emphasizing safeguards for freedom of expression, satire, and political criticism. We wanted to clearly distinguish between content control and process-oriented duties such as transparency, evidence preservation, and rapid response mechanisms. We also worked to diversify our audience—engaging academics, journalists, electoral officials, and civil society—so that the project did not depend on a single institutional champion in a volatile environment. 

The project findings are the result of research, debate, and inputs of a broad and diverse community that are informing the National Electoral Chamber, specialized prosecutors, and provincial electoral courts. We propose the creation of a permanent digital electoral integrity observatory and a multi-actor “Compromiso Ético Digital 2.0” with binding minimum obligations for platforms. 

Regionally, the Argentine case serves as a basis for comparative work with other Latin American countries, contributing to emerging discussions on AI and elections within the Inter-American system and networks of electoral authorities.

Key findings from the project 

• AI-generated deepfakes have become a recurring and strategic tool in Argentine electoral campaigns, particularly effective in late-stage voting periods.
• Accountability gaps persist due to regulatory limitations, slow judicial responses, and the difficulty of attributing responsibility under conditions of plausible deniability. 
• Digital platforms play a central yet insufficiently regulated role, exercising de facto control over information flows without corresponding transparency or accountability obligations.
• Effective responses should prioritize process-based measures—such as transparency, traceability, and rapid coordination—rather than content control. 
• A human rights-based approach, grounded in Inter-American standards, is essential to safeguard both electoral integrity and freedom of expression. 
• Cross-sector collaboration and permanent institutional capacities (e.g., observatories, early warning systems) are critical to move from reactive to preventive governance.

Next steps: Carrying the project forward 

Looking ahead, we believe there is a need to continue working together to build capacity. Combining research, advocacy, and pedagogy enables a community to take a stand against the use of fake news to influence public access to information and the outcomes of public elections.