In his first public event as president of the Superior Electoral Court (TSE), this Wednesday, 20th, minister Kassio Nunes Marques highlighted that managing the Court in a year of general elections, given the advances in artificial intelligence, “is a challenge that will not be small”. According to him, for magistrates, “this is an environment that is still almost completely unknown”.
“Democracy requires trust and faced with the challenges of artificial intelligence, also known as AI, it is up to all of us to ensure that technology is an instrument of citizenship and not of manipulation. Of transparency and not opacity. Of democratic inclusion and not of distortion of the popular will”, said the minister at the opening of the seminar “Seta Debate – Artificial Intelligence in the 2026 Elections”, in Brasília.
For Nunes Marques, digital transformation directly affects the environment where popular will is formed, as processes have become more dynamic and it is possible to spread “technically optimized lies” with the use of AI.
“The voter is no longer reached only as a citizen and is also interpreted as a set of data, presumed preferences, emotional vulnerabilities and reaction probabilities. This is an institutional challenge of enormous relevance”, he stated.
The judge also expressed concern about advances in deepfake videos, that is, content made with AI that realistically imitates a person’s voice, face or movements, facilitating the spread of false information.
“The question is: under what rules, under what controls, for whose benefit, with what degree of transparency this (AI) influence will be exercised”, asked the minister. “If the response is technical, proportional and committed to the dignity of the voter, artificial intelligence can strengthen democracy. If the response is silent or captured by private interests, it could compromise trust in the system.”
Continues after advertising
Nunes Marques also highlighted the importance of bringing together the Electoral Court, experts, academia and civil society to face technological challenges in elections.
“I am here much more to listen, learn and better understand this technological scenario, because the dialogue between Electoral Justice, academia, experts and civil society will be fundamental for the continuous improvement of the Brazilian electoral process”, he stated.
The minister highlighted the technical work carried out by the Court to face possible threats in the digital environment. “Our technical team has been dedicated intensely. I believe we are on the safe path to guarantee harmonious elections, with civility in the democratic environment and also more responsibility on social networks and in the virtual environment”, he said.
For the 2026 elections, the TSE prepared a series of rules on the use of AI in the election with the aim of preventing the production of misleading content during the campaign. The text was reported by Nunes Marques and unanimously approved by the Court.
One of the rules was the ban on the circulation of deepfakes of candidates from 72 hours before until 24 hours after the elections. According to the rapporteur, Nunes Marques, the objective is “to exclude unwanted surprises in the most critical period of the electoral process”. In 2024, the TSE had already banned the publication of this type of content in the electoral context, but had not established a time limit for the ban.
The resolution also provides that digital platforms will have joint liability (they must pay fines or suffer other sanctions) if they do not immediately remove content and accounts during the electoral period that convey materials produced by AI that are not properly labeled or that violate prohibitions.
Continues after advertising
It is also prohibited for AI providers to provide, even if requested by users, candidate recommendations “in order to prevent algorithmic interference in the decision-making process of defining the vote”.
Furthermore, any advertisement that uses an image, voice or content manipulated by artificial intelligence, in printed or digital material, must carry an “explicit, prominent and accessible” warning that it is content produced or altered by AI, in addition to informing which technology was used.
