The Superior Electoral Court’s propaganda resolution () prohibits chatbots such as , Gemini and Grok from recommending votes or favoring candidates in user queries. The text, which will be published by Thursday (5) and to which the Sheet had access, will dictate the rules for propaganda and use of the internet and AI in elections.
The resolution also includes an article that provides for the reversal of the burden of proof in electoral actions regarding the use of synthetic content generated by artificial intelligence in situations where it is very difficult to prove digital manipulation.
If, for example, a candidate or party accuses another of violating the rules for using AI in advertising, but has difficulty demonstrating the irregularity of the content, the electoral judge may reverse the burden of proof. In this case, the author of the publication, together with the platform or not, will have to demonstrate how AI was used and prove the veracity of the information.
In the case of chatbots, when consulted by users, they are prohibited, for example, from prioritizing information about certain candidates to the detriment of others.
to São Paulo city hall in 2024 showed that Gemini, from , when faced with the question “who is” and “who is”, refused to answer. On the other hand, it gave a series of information when the question was asked with the names of the candidates (MDB), (PRTB), José Luiz (PSDB) and Marina Helena (Novo).
According to the new TSE rule, chatbots are prohibited from carrying out any form of electoral favoritism, directly or indirectly. They also cannot make changes to photos or videos that contain nudity or pornography.
“Increasingly, people consult information using chatbots, instead of search engines, and these results can have a huge impact on voters’ decisions; this measure aims to combat imbalance in the dispute”, says Bruno Bioni, founding director of Data Privacy Brasil.
In the case of chatbots, companies providing the applications will need to act proactively, making illegal content unavailable, regardless of legal action, but communicating to users the reason for deleting the content and allowing them to appeal.
The new TSE rule maintains the requirement that all advertising content that uses AI be accompanied by explanations about the use of these resources. Furthermore, the resolution prohibits the publication and republication, free or with promotion, of these synthetic contents in the period between 72 hours before and 24 hours after the end of the election.
An innovation in the TSE rule that will be published in the coming days is the requirement for platforms to have compliance plans to prevent and mitigate risks to the integrity of the electoral process. Platforms are required to demonstrate compliance with each of the resolution’s provisions and publish measurable indicators to monitor their implementation. “For the first time, platforms will have to account for their actions during the electoral process,” says Bioni.
The resolution maintains the prohibition introduced in the 2021 TSE rule on disclosing or sharing facts known to be untrue or seriously out of context that affect the integrity of the electoral process. But, in a nod to right-wing parties, the TSE included a clause protecting freedom of expression, safeguarding “the free expression of the thoughts of an identified or identifiable voter on the internet”.
Following suggestions sent during the public consultation period, the new electoral rule includes a Marçal clause, and now considers competition, ranking or award mechanisms that offer, directly or indirectly, an economic advantage to individuals or legal entities to publish publications of a political-electoral nature to be prohibited. The rule refers to the video cutting championships promoted by then candidate Pablo Marçal in the 2024 election.
