Brazil's Superior Electoral Court (TSE), the body that oversees elections, will rule on Tuesday (September 1) on a lawsuit filed by the Workers' Party (PT) over an AI-generated video that reproduced the image and voice of former president Jair Bolsonaro. The video was shown at the PL party convention that confirmed Senator Flávio Bolsonaro (PL-RJ), the former president's son, as its candidate for president. According to Valdo Cruz's blog at G1, the justices are inclined to keep the ban on deepfakes in the campaign, but are split on whether to fine the candidate.
The PT alleges irregular use of artificial intelligence and illegal early campaigning, according to Congresso em Foco. In the recording, an artificial version of Bolsonaro endorses his son and says, in one passage, that he is "imprisoned and silenced by an unjust decision". The former president is under house arrest after being sentenced to 27 years and three months in prison in the coup plot case, and is barred from publishing statements of a political-electoral nature, including through third parties.
Justices heard by G1 on condition of anonymity point to two camps on the fine. One group favors the penalty, arguing that the material qualifies as a deepfake and could lead part of the public to believe the statement was authentic. The other holds that no sanction applies because the video was shown at a closed party event aimed mainly at convention delegates and party members. Talks are ongoing, and there is no final position on the specific case.
Deepfake ban expected to stand
The general ban in the electoral resolution is expected to be preserved. According to G1, that view holds a majority among the justices and has the support of the TSE president, Kássio Nunes Marques. The central criterion would be the degree of realism: punishable content is that capable of making voters believe something actually happened. Memes, caricatures, satirical montages and animations whose artificial character is evident would be treated differently.
Congresso em Foco reports that Justice André Mendonça proposed this week that the full court adopt this criterion to guide similar cases. According to the outlet, Nunes Marques had earlier floated the idea of allowing a kind of "good AI", used without offensive content and outside negative campaigning, but the interpretation met resistance among his colleagues. The TSE resolution for the 2026 elections prohibits the use, to favor or harm candidacies, of synthetic audio or video content that creates, replaces or alters the image or voice of living, deceased or fictional people. The ban applies even with the authorization of the person depicted.
Defense cites "technological puppet"
Flávio Bolsonaro's defense argues that the use of the technology was not concealed and that the audience was told it was not a physical presence, a live broadcast or an authentic recording. The lawyers compare the device to masks, puppets and other representation tools historically used in political campaigns.
"There was only the creation of a technological puppet, everything with the proper disclaimers and notices, without any falsification or inducement to error, with an appeal that does not come anywhere near an explicit request for votes, asking the delegates to 'welcome with open arms the person I chose to replace me'", the defense said.
Justices heard by G1 say the main focus of the ruling is not the former president's specific situation, but setting parameters that apply to any candidate or party using artificial intelligence in the campaign. The TSE session is scheduled for 7 p.m. on Tuesday (September 1), according to Congresso em Foco. The first round of the elections is set for October 4, with a possible runoff on October 25.