The stakes could not be higher in Brazil’s election. Yet tech platforms are predictably failing to meet the moment.
More than 120 million Brazilians voted in the first round of elections on October 2. A second round on October 30 will determine Brazil’s next president.
This election is likely to be a critical test for democracy and the rule of law in the country, with consequences that go beyond its borders, given Brazil´s size and influence. In recent years, President Jair Bolsonaro, who is running for re-election, has sought to undermine trust in the electoral system, alleging, without providing any proof, that it is unreliable. On October 17, only two weeks before the runoff, Bolsonaro once again questioned the reliability of the electoral system.
As we have seen in elections around the world, privately owned social media platforms and messaging apps have become the de facto public square for campaigning and public debate in Brazil. Tech platforms have a responsibility to respect human rights. This includes the right to participate in democratic elections.
Brazilian civil society warned about the spread election-related disinformation. And in February, the Superior Electoral Court signed memoranda of understanding with Twitter, TikTok, Facebook, WhatsApp, Google, Instagram, YouTube, and Kwai with the aim of stemming disinformation during the electoral process. But the platforms have largely failed to meet their responsibilities. And as predicted, the spread of electoral disinformation risks undermining Brazil’s democratic process.
On election day, and as the vote counting was underway in the following days, posts and videos containing misinformation and allegations of election fraud began to circulate on social media. WhatsApp and Telegram groups with supporters of President Bolsonaro circulated messages saying that “if Bolsonaro didn’t win in the first round, it would be because the elections were fraudulent”. According to the Superior Electoral Court, reports of disinformation increased significantly compared to the first round. The TSE reported receiving 5,869 complaints in the first 11 days of the campaign's resumption, which accounted for almost half of complaints received for the campaign at that point. Complaints grew 1,671% compared to the 2020 municipal election.
People with hundreds of thousands of followers claimed that the counting was fraudulent. This tweet, for example, suggested that the “cheating” had begun on a 15-minute break by the electoral court, insinuating that it would favor Bolsonaro’s opponent. The tweet is still online, has over 30 thousand likes, and lacks any link to information from the electoral authorities, the official results of the election, or other accurate information. Other tweets remain online that explicitly claim voter "fraud' without any credible information, and do not contain a label or link to authoritative information.
There are three key dimensions to this problem:
First, Brazilian civil society groups have noted that no platform besides Twitter has a policy to prevent calls for insurrection against the democratic order or interference in the peaceful transmission of power that do not explicitly call for violence. This means that platforms could be used to organize and promote anti-democratic actions in the event that there’s an institutional crisis following the election.
Second, when it comes to electoral fraud, platforms have widely different policies, and even the stronger ones aren’t always enforced. Meta’s publicly available policies, for instance, prohibit the promotion of paid content that alleges electoral fraud but don’t address similar unpaid content.
A survey by Federal University of Rio de Janeiro’s Netlab found that Meta allows some content to circulate on Facebook and Instagram, even after being classified as electoral disinformation by independent fact-checkers, in some cases without labels. Most of the posts in the study contended that electronic voting machines are not reliable and that it would be easy to rig the election. The study also found that the profiles of those who post the most electoral disinformation were politicians.
Another Netlab study documented at least four cases of paid ads on Meta services containing unfounded claims that questioned the reliability of the vote count.
Twitter has a policy on civic and electoral integrity. But as shown by the tweet repeating false claims of electoral fraud the company does not always adequately apply the policy. YouTube’s electoral integrity policy is restricted to false allegations of voter fraud in previous elections.
An investigation by data analysis consultancy Novelo Data published by Folha de São Paulo found that from October 2-15, 16 livestreams and 137 videos were posted on YouTube with allegations of fraud without evidence. In all, the content had at least 3.3 million views. Much of this content is still available. Telegram is an outlier. It has no published policy to address disinformation and attacks on democracy, and it has failed to act on its March commitments. Telegram and WhatsApp are reported to be the main platforms used to disseminate disinformation about the electoral process, with groups and channels with tens and thousands of followers dedicated to spreading this narrative. Closed messaging platforms pose distinct challenges to detecting and curtailing disinformation.
Third, tech platforms have a track record of bending their rules for powerful actors, and allowing politicians to get away with speech that violates their policies. Access to what politicians say is indeed crucial to holding leaders accountable. But being more permissive of powerful actors can allow them to incite violence on these platforms, or cause other harm, with few consequences.
Since politicians have more influence in society than ordinary people, their statements have, in fact, much more potential to lead to harm. Many of the biggest disseminators of fake news in Brazil are at the highest level of national politics. Much of the distrust in the Brazilian electoral system was propagated by President Jair Bolsonaro and his sons, who are also politicians.
Human Rights Watch wrote to Google/YouTube, Meta, Telegram, and Twitter to inquire about their efforts to mitigate electoral disinformation in Brazil’s election. No company directly addressed our specific questions, however Meta said that it had created tools that promote reliable information and label election-related posts, established a direct channel for the Superior Electoral Court to send it potentially harmful content for review, provides access to data about content removed from Facebook and Instagram during the campaign period, and continues closely collaborating with Brazilian authorities and researchers.
Twitter's written response pointed to its Civic Integrity policy, which addresses misleading information that could impact the integrity of civic processes, says it will remove and/or add labels and context to misleading, harmful Tweets and reduce their visibility. Twitter notes that opinions and different points of view and interpretations are not necessarily actionable under its policy.
To meet their responsibilities, platforms should promptly comply with orders by the Superior Electoral Court to remove electoral disinformation. They should also fill gaps in their policies and enforce them to ensure that they are respecting the right of Brazilians to participate in democratic elections. This requires adequately resourcing efforts to protect election integrity and civic discourse and being transparent with and accountable for their actions. Platforms should also hold politicians to a higher standard when it comes to speech that is likely to incite violence or spread harmful disinformation that could undermine the democratic process.
Targeted efforts to address electoral disinformation are necessary but insufficient. More broadly, platforms need to address their chronic underinvestment in user safety around the world. Regulators should address the rights abuses facilitated by the underlying business models of dominant platforms, which rely on the massive collection of personal data to sell access to people’s attention and are largely designed to prioritize their “engagement” over human rights.
* A version of this article has been published at UOL.