President Luiz Inácio Lula da Silva’s campaign asked Brazil’s Superior Electoral Court on Monday to investigate Meta, accusing the company of manipulating its algorithms to help Senator Flávio Bolsonaro six days before Sunday’s presidential runoff.
The filing from Mr. Lula’s Workers’ Party says Meta’s recommendation systems, generative AI tools and ad review worked for Mr. Bolsonaro and against the president, and that the company ignored an earlier court order, O Tempo reported. Meta owns Facebook, Instagram and WhatsApp. Mr. Bolsonaro, the eldest son of former President Jair Bolsonaro, led Mr. Lula in the first round.
“We hope Meta stops benefiting the Bolsonaro side’s campaign and behaves ethically,” said Éden Valadares, the party’s communications secretary.
The court has already stepped in. On Oct. 15, Justice Marco Aurélio Bellizze barred Meta from restricting the delivery of posts about presidential candidates, and on Monday he gave the company 24 hours to prove it was treating the two campaigns equally. Meta sent the court a report on Tuesday describing the steps it had taken.
Meta denies favoring any candidate. It has blamed technical errors for incidents like the removal of Mr. Lula’s ads, Umar Farooq reported for Al Jazeera. The president’s own ad account was briefly suspended last month.
His campaign has been collecting complaints since Oct. 14, when it opened a website asking supporters to report blocked or throttled posts. Its petitions lean on outside research.
Daniela da Silva of CTRL+Z, a group tracking such cases, said campaign organizers watched their organic reach drop and had to pay to boost posts. Some reported spending more than 70 percent of their boost budgets on Meta. The group has logged 1,100 reports this year of takedowns or restricted reach, “mainly related to profiles supporting Lula,” Ms. da Silva said.
The money is large. Meta took in at least 288 million reais ($57 million) in direct campaign spending this year, twice the 2022 total, according to an analysis by the Brazilian news site G1. More than 76 percent of Brazilian voters now get their news online, said Andre K. Rodarte of the Digital Democracy Centre at the University of Southern Denmark.
Other researchers have tested Meta’s ad screening. The advocacy group Eko submitted 15 election ads containing AI-generated violent content and disinformation, and Meta approved 13. Meta said it had removed ads that broke its policies and that the small sample didn’t reflect the scale of its moderation.
NetLab, at the Federal University of Rio de Janeiro, found more than 1,600 political ads paid for from abroad that were apparently labeled as nonpolitical. They brought Meta 2 million reais ($400,000). Some offered online voting, which Brazil doesn’t allow, through a site that harvested personal data. “The damage was already done,” said Debora Gomes Salles, NetLab’s head of research.
Brazil tightened its rules after Jair Bolsonaro’s supporters stormed Congress, the Supreme Court and the presidential palace on Jan. 8, 2023. Deepfakes of candidates are banned, political ads must disclose AI use, and only officially backed campaigns may run ads. Vicky Wyatt of Eko said the riots “really shocked a lot of policymakers and political elites to act.” Meta has said the rules “increase the risk of over-enforcement and threaten free expression.”
The elder Mr. Bolsonaro is barred from office for eight years, serving a 27-year sentence for plotting a coup and banned from social media. Leticia Cesarino, an anthropologist at the Federal University of Santa Catarina, said the left still trails the far right online because the attention economy rewards “alarmistic, simplistic, affect-laden far-right discourse.”
Meta’s own Oversight Board cited “recurring” failures in the company’s moderation of election content on Tuesday and called for clearer, more consistent rules. Brazilians vote on Sunday.