Paragraph 1
In 2026, Brazil will hold its first elections where artificial intelligence is not just a novel tool but a widespread, easily accessible feature of the digital landscape. This evolution brings with it a chilling prospect for researchers and democracy advocates: a massive amplification of the disinformation challenges that have already plagued recent elections. Débora Salles, the general coordinator of NetLab, a research laboratory at the Federal University of Rio de Janeiro, warns that while the problems of 2022 and 2024 persist, they now come with a “boost” of scale and reduced cost, thanks to AI. The central concern is that the electoral justice system is not prepared to enforce existing rules, and the platforms themselves have little incentive to act proactively. The result is an environment where foreign interference, undetected political ads, and sophisticated deepfakes can flourish, undermining the very foundation of a free and fair vote. The promise that technology would connect and inform citizens now feels overshadowed by its potential to deceive and polarize, setting the stage for an electoral process fraught with unprecedented risk.
Paragraph 2
The core of the issue, according to Salles, lies in the murky world of online political advertising. The scale of the problem is so vast and the lack of transparency from platforms so profound that meaningful oversight is nearly impossible. NetLab’s research has repeatedly uncovered illegal advertisements that, while prohibited, were relatively easy to find. In 2024, her team identified 137 such ads, many traced back to accounts registered in foreign countries, including Portugal, the UK, and Belgium. Salles points to a specific, alarming case where ads promoting the candidacy of São Paulo mayoral candidate Pablo Marçal were run through TikTok pages based outside of Brazil, across Europe, Asia, and the US. This highlights a critical vulnerability: the platforms’ verification systems are woefully outdated, allowing anyone to exploit loopholes. Despite platforms publicly claiming to prohibit foreign election advertising, there is no functional mechanism to stop it. The inability to trace the source of these ads makes it impossible to measure the extent of foreign interference, leaving the election open to malicious actors with impunity.
Paragraph 3
If the current environment is already problematic, Salles argues that artificial intelligence is poised to pour gasoline on the fire. The dramatic reduction in the cost and effort required to produce compelling synthetic content is a game-changer. What once required technical expertise is now trivial for the average user, and this content gains significant traction online, often outperforming factual information. Salles explains that this creates a perfect storm: the platforms’ algorithms are designed to maximize engagement, and AI-generated sensationalism is highly engaging. The platforms, which are financially incentivized to show users more content, are therefore complicit in amplifying the very misinformation that threatens the electoral process. This cycle of creation and propagation is devastating because it pollutes the public square, leaving ordinary citizens to sift through a landscape where truth is difficult to distinguish from fabrication and where the integrity of the entire election is brought into question.
Paragraph 4
A concrete example of the concerning intersection of politics and AI occurred at a convention for Flávio Bolsonaro, where a deepfake video of former President Jair Bolsonaro was shown. Salles is unequivocal: such content is a deliberate violation of the Superior Electoral Court’s (TSE) rules. She clarifies that the resolution is explicit that fabricated or manipulated content cannot be used to benefit a candidacy or harm another. While AI can be used for innocuous purposes like enhancing a background or adjusting one’s physical appearance in a photo, creating a realistic video of a real person discussing a topic is forbidden in the context of the campaign. However, the existence of clear rules does not translate into effective enforcement. The incident highlights the massive gap between policy and practice, where the very people who should be finding solutions are perpetuating the problem. The question looms: if clear violations of the rules are not swiftly punished, what precedent does that set for the rest of the campaign?
Paragraph 5
The responsibility for mitigating these risks falls to the TSE, whose efforts Salles views with a mix of hope and deep skepticism. While the TSE has engaged in meetings with tech companies and issued a guide for best practices, Salles argues that the measures are insufficient because they do not hold the platforms accountable. She criticizes the approach as “almost generous” for relying on cooperation rather than obligation. When the TSE does act, enforcement is slow and fines are too low to serve as a deterrent. This lack of precedent, Salles suggests, is partly due to the novelty of the situation; these are the first major elections where AI is so fully integrated into daily life. However, she also points to the overwhelming lobbying power of technology companies, who are resistant to any regulation that might impede their profitability. The TSE’s recent initiative to partner with external organizations to monitor synthetic content is a step in the right direction, but it may be too little, too late. These systems are still rudimentary, and building a nationwide monitoring infrastructure from scratch just months before the election is a herculean task that requires more time and resources than are being allocated.
Paragraph 6
In the face of such systemic challenges, Salles insists that the ultimate safeguard lies with the public. She cautions against placing the burden on individuals, but stresses that voters must remain vigilant—checking multiple credible sources, questioning information that elicits an emotional reaction, and understanding that if something is designed to provoke, it is often designed to distract. Reporting suspicious content to the TSE and civil society organizations is a critical, yet underutilized tool. The mountain of data is immense, and no institution can go it alone; society’s collective help is the only viable bulwark against a landslide of disinformation. Salles concludes with a stark assessment: we are facing a challenge of enormous proportions, and it is clear that neither the TSE nor the public is entirely prepared. The fight for the integrity of the 2026 election will not be won in a single moment but through a constant, collective effort to demand transparency and accuracy. The coming months will be a testament to whether these mechanisms can be built in time, or whether democracy itself becomes the latest casualty of the algorithm.