Meta Denies Brazil Election-Bias Claims — SkimNews
Get the Geopolitics newsletter
Daily geopolitics — wars, elections, sanctions, the diplomatic moves that move markets. Free.
- Lula’s campaign formally accused Meta of favoring Flávio Bolsonaro by suspending the campaign’s advertising account, suppressing posts and blocking AI-generated content ahead of the 25 October runoff.
- Brazil’s electoral prosecutors said Meta failed to fully comply with court orders requiring equal treatment of Lula and Flávio; UN human rights experts raised the same concern, while Meta’s independent oversight board said election safeguards had failed.
- Meta called claims that it was favoring or interfering for candidates or political groups false and unfounded; it said Lula’s account was suspended for one hour by mistake and its systems blocked prompts referring to “Lula” and “Bolsonaro,” but not “Luiz” or “Flávio.”
- Meta AI, the chatbot integrated into WhatsApp, appeared in circulated images refusing to generate Lula campaign posters while instantly producing one for Flávio; screenshots submitted by Lula’s campaign also showed it repeatedly assigning Bolsonaro’s electoral number 22 to fictional politicians and allegedly translating “Flávio future prisoner” as “Flávio future president.”
- Lupa identified 920 pieces of AI-generated or manipulated election content before the first round, primarily on Facebook and Instagram, with 60% unlabelled.
- Meta advertising accounts numbered 26 and appeared to sidestep restrictions on foreign-funded political advertising by posting AI-generated videos, including one depicting Flávio Bolsonaro and Argentina’s president Javier Milei freeing Brazil from communists.
- InternetLab found that eight major AI chatbots sometimes refused genuine candidate questions and referred users to official electoral websites, but agreed to draft misleading claims questioning the integrity of elections for campaigning.
Why it matters: Meta’s handling of Brazil’s 25 October runoff sits under scrutiny from electoral prosecutors, UN human rights experts and its independent oversight board after the company failed to fully comply with equal-treatment orders. The stakes are concrete: 920 AI-generated or manipulated election items were catalogued before the first round, 60% unlabelled, alongside 26 advertising accounts that appeared to bypass foreign-funding limits.
Ask SkimNews
