São Paulo –Days after Meta disabled President Lula's official campaign ad account, new research from corporate accountability group Ekō shows the platform is profiting from shadow accounts that push AI deepfakes, conspiracy theories and election disinformation to Brazilian voters. The research comes as Brazil’s voters prepare to go to the polls this Sunday.
Researchers identified 26 ad accounts that together spent $223K (R$1.1M) on Meta ads, many of them not authorised to run electoral ads at all. All of the ads analysed appear to breach both Meta's rules and Brazilian electoral law.
The ads include an AI-generated video of Flávio Bolsonaro and Javier Milei on a "mission" to "free Brazil from communists." Others push attacks on Brazil’s Superior Electoral Court and raise doubts about the integrity of Brazil’s electoral system. The accounts analysed in this research posted over 3000 ads in 90 days.
Political advertising in Brazil is tightly regulated by the Brazilian Superior Electoral Tribunal (TSE). Only candidates, political parties and official campaigns are permitted to pay for electoral ads. Many of the accounts appear to be shadow accounts that pose as news outlets or fan pages to get around Brazil's election rules. These shadow accounts are able to run attack ads, dodge AI labelling requirements, and publish ads from abroad using foreign funding.
Most of the ads ran under Meta's "Social Issues, Elections or Politics" category, which is meant to be subject to stricter review. Meta removed many of them, but often only after they had been live and had gathered thousands of impressions.
Vicky Wyatt, Ekō Campaign Director, said: "Meta cut off the sitting president's campaign ads without explanation. At the same time, it's taking money from dodgy accounts, some run from outside Brazil to flood voters' feeds with deepfakes and election lies. Meta has every tool it needs to enforce Brazil's election rules. Instead, it keeps cashing in at the expense of Brazil's democracy."
"Brazil's authorities deserve to know why Meta is letting these ads run at full force, and they need answers before it is too late.."
This is Ekō's third investigation into Meta's role in Brazil's election. The first found mainstream AI chatbots recommending and amplifying election disinformation to voters. The second showed Meta's ad review system approving AI-generated hate speech and disinformation. This report shows the same failures already live in Meta's public Ad Library.
Brazil votes in the first round of its presidential election on Sunday, 4 October.
ENDS
For further information please contact press@eko.org
For more information about a current Ekō campaign or to arrange an interview with a spokesperson please contact us at