Brazil’s Superior Electoral Court (TSE) will begin hearing a case Tuesday that could establish how artificial intelligence may be used in the country’s 2026 election campaigns. The session is scheduled to start at 7 p.m. Brasília time.
The case was brought by the Workers’ Party (PT), part of the Brazil of Hope federation with the Communist Party of Brazil (PCdoB) and the Green Party (PV). It concerns a video produced with AI by former President Jair Bolsonaro and shown at the national convention of the Liberal Party (PL), where Senator Flávio Bolsonaro was formally nominated for president.
CNN Brasil is the sole source for this report.
The Video at Issue
The recording shows an avatar with an image and voice resembling Jair Bolsonaro. In the video, the avatar says he is under arrest because of what he describes as an unjust decision and endorses his son’s candidacy.
Bolsonaro is under house arrest and, by an order from Supreme Federal Court (STF) Justice Alexandre de Moraes, is barred from making public political statements during the election period. The PT argues that the video amounts to early election advertising and an unlawful use of artificial intelligence.
Flávio Bolsonaro’s defense says the video should be treated like campaign devices such as masks or puppets: an obviously artificial representation that was not intended to deceive voters.
Synthetic Content or Deepfake?
Brazil’s current electoral rules require campaign material produced or manipulated with AI to be explicitly identified to voters. They also prohibit deepfakes used to benefit or harm candidates.
The unresolved question is when AI-generated material stops being permitted synthetic content and becomes an illegal deepfake. The distinction could affect campaign videos, political satire and other forms of digital communication during the election.
Justice André Mendonça recently proposed separating the categories according to how realistic the material appears. Under his approach, synthetic content would include memes, caricatures, animations and satire whose artificial nature is clear. Such material could be used if it is properly identified.
Deepfakes would be limited to realistic content objectively capable of making voters believe they were seeing an authentic recording. Their use would remain prohibited whether the material favored or damaged a candidate. Mendonça cited the European Union’s AI Act as a reference for the standard.
A Divided Court
The TSE had already been debating the issue internally. A private meeting in August ended without consensus after the Bolsonaro video drew attention.
TSE President Kassio Nunes Marques supported allowing AI uses that were neither offensive nor part of negative campaigning. The idea became informally known as “AI for good.” Other justices resisted, arguing that the electoral rules prohibit deepfakes used for either positive or negative purposes, regardless of intent.
Mendonça’s proposal focuses less on the purpose of the content and more on its ability to appear real. Beyond deciding whether the PL convention video violated electoral rules, the court is expected to indicate which interpretation will guide future cases and may impose a fine.