On October 9, Meta’s own engineers spent nearly 24 hours unable to fix a problem that was stopping Brazilian President Luiz Inácio Lula da Silva’s campaign from boosting posts on his Instagram feed. In that same window, according to internal messages reviewed by newspaper Folha de S.Paulo, Instagram’s recommendation system was independently pushing out the profile of his runoff opponent, Senator Flávio Bolsonaro, along with Bolsonaro’s WhatsApp channel — the kind of algorithmic boost Brazil’s electoral court has explicitly barred during the race.
Reuters also reported Friday on the existence of these messages, mentioning that Meta employees acknowleded internally, that the Instagram accounts of Lula were negatively affected by restrictions during the country’s election, while the platform recommended the profile of challenger Flavio Bolsonaro, as well as his WhatsApp channel.
None of the internal conversations that Folha accessed contain an admission of deliberate interference. However, the staff did discuss the severity of the bug and how slowly engineering moved to fix it. One employee flagged that the issue needed to be classified at the company’s highest urgency tier, warning colleagues it involved the sitting president’s account in the middle of a campaign. Another worried aloud about how the company would eventually explain the outage to members of Lula’s team who were already asking questions, while a colleague reported that lawmakers who’d joined Lula’s campaign were pressing for answers of their own.
As for Meta, the company had its own technical explanation internally: staff reportedly attributed Lula’s boosting failure to his name being too long for a backend system, with the bug supposedly fixed, while calling the erroneous promotion of Bolsonaro’s profile a separate, unrelated mistake. But Brazil’s electoral rules were written to police outcomes, not intent, and a nearly day-long failure that happened to suppress the incumbent while a second glitch happened to elevate his opponent is exactly the kind of asymmetry those rules exist to catch.
When Lula’s campaign took its complaints to the Superior Electoral Court, it didn’t mince words, arguing that “profiles acting as political operatives for the opposing side … receive a disproportionate and artificial reach, garnering more views even than global celebrities.” Meta’s public response to the court, filed October 11, rejected any suggestion of bias while admitting it had faced technical glitches that affect accounts beyond Brazil. That response hasn’t aged well against what Folha’s messages now show happening behind the scenes. By October 15, a judge had ruled that Meta must stop any practice that creates an imbalance between the two candidates, whatever the cause.
This hasn’t been Lula’s first run-in with Meta’s moderation systems. Earlier in the race, the company had briefly pulled Lula’s ability to run political ads over unspecified rule violations and separately took down an Instagram video examining Bolsonaro’s ties to a banking fraud case — both reversed within hours after public pressure, both never fully explained. The pattern predates Friday’s revelation; Folha’s reporting just supplies the first documentation of what Meta’s own staff were saying about it internally.
The same internal messages reportedly describe Bolsonaro’s campaign having a batch of ads blocked on October 1 after Meta’s systems flagged AI-generated material inside a window when the electoral court bans its use — a restriction his campaign disputes. Meta had not responded publicly to Folha’s report as of Friday. With the runoff nine days away and both sides already in court, the company is running out of time to decide whether “technical issue” is still the explanation it wants to stand behind.