As Brazilians prepare to return to the polls on 25 October, the atmosphere is electric and uneasy. The first round on 4 October left the nation split down the middle: Flávio Bolsonaro of the Liberal Party captured 47% of the vote, while incumbent Luiz Inácio Lula da Silva of the Workers’ Party took 45%. That razor-thin margin of less than two points, combined with a shortened campaign window and a deeply polarized electorate, is a recipe for chaos—especially in the digital realm where disinformation thrives. The stakes could not be higher, and everyone knows it. The political discourse has moved far beyond simple policy debates; it is now a battle over the very integrity of the electoral process itself. In this hyper-connected age, the battlefield is not just the ballot box but also the screens of millions of Brazilians who are bombarded daily with a flood of information, much of it manipulated, synthetic, or outright false. The country is experiencing a pivotal moment that will test the resilience of its democratic institutions against the relentless march of technology and the shadowy forces that seek to exploit it.
At the heart of this struggle is Brazil’s Superior Electoral Court (TSE), which has become the de facto regulator of the digital electoral environment. While the court sets the rules for online political speech, it works in tandem with the platforms—Facebook, WhatsApp, TikTok, and others—that implement and enforce those rules. This partnership is supposed to safeguard electoral integrity, but the past decade has shown just how fragile that protection can be. Ever since the 2016 US presidential election and the Brexit referendum, we have seen how disinformation campaigns can exploit digital infrastructure on a massive scale. Brazil’s own 2018 election was a grim preview, with WhatsApp becoming the central conduit for coordinated lies. In response, the TSE has tried to stay ahead of the curve, issuing resolutions that establish boundaries for advertising, technology misuse, transparency, and cooperation with digital platforms. But the landscape is shifting faster than any rulebook can keep up with. What started as a problem of social media manipulation has now morphed into a far more complex crisis involving generative AI, foreign interference, and opaque platform governance. The first round of this election has already revealed the new challenges: AI makes disinformation cheaper, faster, and easier to produce, while the court and the platforms are scrambling to catch up.
To address these challenges, the TSE has taken unprecedented steps. In 2024, it issued Resolution No. 23,732/2024 to regulate AI use in municipal elections, explicitly banning deepfakes and manipulated content. For the 2026 general elections, Resolution No. 23,755/2026 went much further, holding platforms directly liable if they fail to remove illegal synthetic content or unlabelled deepfakes after being notified by the court. The new rules also mandate immediate removal of content attacking electronic voting machines or promoting anti-democratic acts, without waiting for a court order. There is even a strict blackout period: no new synthetic media featuring candidates or public figures can be published from 72 hours before the election until 24 hours after. Platforms are required to label all AI-generated campaign materials, maintain transparent ad repositories, and submit compliance plans. On paper, these are powerful safeguards. And the numbers seem to reflect progress: TSE data shows an 86% decrease in electoral disinformation alerts compared to 2022, falling from 18,487 to just 2,604 between August and the eve of the vote. But specialists are not convinced. They argue that the decrease may be due to early intervention and better cooperation, but the nature of the disinformation has simply changed. Instead of attacking voting machines, the new narratives are more episodic, focusing on alleged collusion between the government, the Supreme Federal Court, and external actors—claims that remain unproven but can still sow doubt.
The rise of generative AI has made everything more complicated. Brazil is now one of the largest markets for ChatGPT, with users sending about 215 million messages a day. This widespread use means that AI tools are not just novelties; they are active participants in the political conversation. A study by ITS-Rio and the Federal Public Prosecutor’s Office found that popular AI tools recommended candidates in 20% of responses, with Perplexity being the most eager (89%), followed by Meta AI, Grok, and DeepSeek. Some chatbots even endorsed specific contenders or questioned election results. Another study by InternetLab and FGV showed that Gemini ranked candidates in 64% of cases and agreed with user biases in nearly 80% of voting chats. Worse, AI has been used to create synthetic voters on TikTok—videos of fake people declaring support for Lula or Bolsonaro that amassed over 7.4 million views. These AI-generated personas are designed to exploit stereotypes and attack opponents, often evading real-time moderation. Even when platforms try to regulate, their policies are often opaque, making it difficult for users, researchers, and candidates to understand what is allowed or how to contest decisions. The result is a widening information gap between the platforms that control AI and the public that relies on them for truthful information.
Platform accountability has been a major sore point. Meta, the parent company of Facebook and WhatsApp, has faced intense scrutiny for its enforcement actions. In the week before the election, Meta briefly blocked Lula’s campaign page from running political ads, citing unspecified rule violations, then lifted the restriction after an appeal—without explaining the reason. Meta also took down legitimate journalistic content, prompting accusations of censorship from digital rights organizations. Researchers uncovered hundreds of irregular ads on Meta and Kwai during the mandatory blackout period, including some that falsely simulated official TSE voter instructions. The blackout period for AI-generated content even forced platforms to delete informative posts by mistake. Meanwhile, transnational interference has added another layer of complexity. Civil society groups traced a coordinated campaign originating from South Korean right-wing networks that amplified claims of Brazilian electoral fraud, generating millions of views. While the Organization of American States (OAS) found no concrete evidence of foreign interference, the fear persists—not just in Brazil but across the Americas, where heads of state increasingly comment on other countries’ elections. The judicial system is also under strain, with single-judge rulings from the Supreme Federal Court and TSE ministers raising concerns about legal uncertainty and inconsistent standards. The TSE president himself warned that overlapping competencies risked fragmenting control over the electoral process.
As the runoff approaches, the world is watching. Brazil is not just fighting for its own democratic stability; it is serving as a testing ground for the rest of the globe. In the coming half-year, 47 elections will take place worldwide, including in Israel and the United States. What happens on 25 October will send ripples far beyond Brazil’s borders. The past months have shown that even with robust intervention from the TSE and regulatory safeguards, both the court and digital platforms face persistent challenges in preserving electoral integrity while protecting free expression. The final round will be the ultimate test of whether they can protect democratic stability in a hyper-polarized environment where disinformation and emerging technologies intersect, creating rapid and novel threats. For the millions of Brazilians heading to the polls, the choice is not just between two candidates; it is between trusting the democratic process and succumbing to the chaos of manipulation. The future of democracy in the digital age hangs in the balance.