Brazil's electoral court gives Meta 24 hours over alleged pro-Bolsonaro bias in searches

Fresh evidence filed with Brazil's electoral court

Two Brazilian organizations, Sleeping Giants Brasil and CTRL+Z, have submitted new evidence to the Superior Electoral Court alleging that Instagram and WhatsApp keep producing results that favor the presidential campaign of Senator Flávio Bolsonaro over President Luiz Inácio Lula da Silva, despite a standing court order requiring equal treatment of both candidates.

The filing was first reported by journalist Mônica Bergamo in Folha de S.Paulo, with further detail from a Brasil 247 report translated by Panamerican Dispatch. The petition is addressed to Justice Marco Aurélio Bellizze, who oversees the case. It asks him to order corrections within 24 hours, to require Meta to explain how it ranks search results and generates answers with artificial intelligence, and to enforce a daily fine of R$500,000 (about $99,000) the court has already set.

What the evidence shows

The new records concern how Meta's automated systems organize and surface political content.

According to the report, a WhatsApp GIF search for "brasil" returned an image of Flávio Bolsonaro as the first result, showing his ticket number and running mate. The same result appeared on three separate accounts, which the groups cite to question whether its placement comes from systematic ranking criteria. A similar check on Instagram placed three images of Bolsonaro among the top results, while the only appearance of Lula was his name in writing, with no photo.

Other examples tie Lula to derogatory content. A WhatsApp GIF search for "ladrão" (thief) returned a caricature of the president in a striped uniform referencing prison dress, alongside a result reading "Supporting the Workers' Party is a mental illness." An Instagram search for the same word surfaced a GIF reading "Vote 13," the Workers' Party ballot number tied to Lula.

The petition also challenges hashtag ordering on Instagram. A search for "lula" reportedly offered "#lulaladrão" as the first suggestion, while a search for "bolsonaro" placed the critical "#bolsonaronuncamais" among its last suggestions.

Meta AI drawn into the dispute

The organizations also tested Meta AI, the assistant built into WhatsApp. Two people asked it to pick a number between 1 and 25, with no instructions and no reference to the election. One answer was 13 and the other was 22 — the ballot numbers of Lula and Flávio Bolsonaro respectively. The groups question the coincidence and want the system's parameters explained, while conceding that the test alone does not show electoral preference.

That framing runs through the whole filing: the organizations say they have not proven bias, arguing instead that the patterns justify court-ordered scrutiny of the relevance criteria behind image rankings, hashtag suggestions and AI-generated answers.

Court raises the pressure

The filing arrives as the electoral court escalates its stance toward Meta. Justice Bellizze had already ordered the company to apply identical criteria to both campaigns and to prove it was complying. On Friday the full court upheld his decision, and on Sunday he gave Meta a new 24-hour deadline to show it had met part of the requirements. He also warned that the daily fine could rise from R$500,000 to R$1 million (about $198,000) if noncompliance continues.

The case began with complaints from Lula's campaign about limits on the circulation of content about his candidacy, and differences in how Meta's tools behaved compared with tools tied to his opponent.

Meta has denied deliberately favoring any candidate and previously blamed the problems on technical failures. Folha de S.Paulo contacted the company about the new records, but it had not responded by publication time, leaving the mechanisms behind the reported results unexplained.

Separately, Sleeping Giants Brasil and CTRL+Z planned a peaceful protest called "Meta, Hands Off Our Election" outside the company's São Paulo headquarters on Monday evening, intending to hand a notice to Meta's Brazil director.

Why it matters

The dispute is not about whether partisan content exists on Meta's platforms, but about how automated systems rank and present it during a presidential race. Brazil's electoral court is effectively demanding algorithmic transparency under threat of escalating fines, testing whether judicial orders can reach into recommendation engines and AI assistants. A ruling that compels disclosure of ranking and AI criteria would set a precedent for election regulation far beyond Brazil. It also puts Meta AI under a new kind of scrutiny, treating its outputs not as a neutral novelty but as part of the informational infrastructure of an election.