The ruling comes ahead of the runoff
Brazil’s Superior Electoral Court ordered Meta to apply equivalent criteria to content and artificial intelligence functions concerning Luiz Inácio Lula da Silva and Flávio Bolsonaro, ahead of the presidential runoff on October 25. The measure was issued after an appeal by Lula’s campaign, which accused Instagram, Facebook and WhatsApp of applying Meta AI algorithms and restrictions in a way favorable to the opposing candidate.
The decision does not automatically certify that Meta manipulated the election. It establishes obligations for the platform during a politically sensitive phase and requires equal treatment as described by the Associated Press. The campaign’s accusation and the company’s explanations are opposing positions; the judge’s order, by contrast, is an act the platform must comply with.
Meta disputes the political interpretation
Meta denied favoring a candidate. The company attributed some reported incidents to global technical problems, anti-fraud measures and rules applied to comply with regulations. Among the examples cited by the company are an interruption in displaying Reels counts, which it said affected more than one hundred countries without interrupting content distribution, and a temporary problem in creating ads from Instagram posts.
On the image-generation function, Meta acknowledged an anomaly: the use of common names together with election numbers reportedly temporarily blocked some images of Lula, while for a period it was still possible to generate images of Flávio Bolsonaro. The company maintained that this was a technical flaw and stated that the rules are applied in the same way to candidates and political orientations.
The most delicate point concerns the distinction between an isolated malfunction and a systemic effect. An error can affect part of a campaign without having been designed to do so; but if the result alters visibility or access to tools during an election, the democratic consequence can be concrete even in the absence of political intent.
Algorithm transparency remains difficult to verify
Platforms control data and recommendation systems that are not fully visible to candidates, researchers or authorities. Verifying equality requires public criteria, logs of interventions, access to comparable data and rapid procedures for contesting errors. Without this evidence, both the accusation of favoritism and the company’s denial are difficult to assess from the outside.
The Brazilian case thus places at the center the ability of institutions to intervene before a campaign ends. The court order is an immediate remedy; verification of effects, transparency of rules and accountability of platforms remain broader questions that do not end with the vote.
