Brazil's top electoral court clarifies AI deepfake rules in political ads

Brazil's highest electoral authority has established a formal definition for artificial intelligence-generated deepfakes in political advertising, a decision aimed at curbing misinformation ahead of the October presidential elections.

The ruling, delivered late Tuesday, came in response to a case involving opposition Senator Flávio Bolsonaro. During his party's convention in July, the senator used an AI-generated image of his father, former President Jair Bolsonaro, who is currently imprisoned. The synthetic depiction appeared to endorse the senator's campaign against President Luiz Inácio Lula da Silva.

The former president was sentenced in November to 27 years in prison for his role in leading a coup attempt.

In a 4-3 vote, the court declined to penalize Senator Bolsonaro, determining that the content was created for a private party event before his candidacy was formally announced, rather than for public broadcast as a political advertisement.

However, in a separate 5-2 decision, the court defined a deepfake in political advertising as "synthetic content produced or manipulated via artificial intelligence or equivalent technology" that possesses a degree of realism sufficient "to create, reproduce or alter the image, voice or speech of any living, deceased or fictional person."

This new standard will apply to a broad range of candidate content leading up to the first round of voting on October 4. While online platforms play a significant role in Brazilian elections, traditional television and radio advertising remain highly influential. Broadcasters are required to allocate 14 minutes daily for nearly 1,000 30-second ads, in addition to dedicated 12-and-a-half-minute blocks reserved exclusively for presidential candidates.