Brazil's Electoral Court Clears Flávio Bolsonaro of AI Video Penalties and Sets Deepfake Criteria

The Superior Electoral Court (TSE) voted 4 to 3 on Tuesday to reject a request to fine presidential candidate Flávio Bolsonaro for using an artificial intelligence avatar of former President Jair Bolsonaro during a party convention in July. The court majority, led by President and rapporteur Kassio Nunes Marques, determined that the video did not constitute irregular early electoral propaganda. The decision followed a challenge by the Brazil Hope Federation, which argued the material violated rules prohibiting deepfakes used to favor or prejudice candidacies.

Beyond the specific case, the TSE established a formal legal thesis to guide the application of rules regarding deepfakes in elections. By a 5 to 2 vote, the court defined an illegal deepfake as synthetic content produced by AI that possesses a degree of realism or verisimilitude and is intended to create or alter the image, voice, or manifestation of a person. The court further ruled that the prohibition of deepfakes only applies when the content is characterized as electoral propaganda. Additionally, the court decided 4 to 3 that the transmission of a party convention via social media does not automatically turn a manifestation of political support into early electoral propaganda; instead, the nature of the act must be analyzed based on context, language, and the intended recipient.

Justice Nunes Marques stated that the rules should be interpreted based on freedom of expression and that the goal is to protect voters from false perceptions of reality. Conversely, dissenting justices, including Ricardo Villas Bôas Cueva, argued that the prohibition should be objective and not depend on the degree of realism. Justice Cueva contended that the video became illicit once it was broadcast live on social networks.