Brazil’s Electoral Court Justice Sets Deepfake Criteria and Finds No Illegality in Bolsonaro AI Video

Brazil’s top electoral court is drawing a line between convincing synthetic media and less realistic forms of digital alteration. TSE President Kássio Nunes Marques voted to define three elements of an illegal deepfake during an election campaign: the content must be synthetic, highly realistic and capable of confusing voters, and intended to create, reproduce or alter the image or voice of a living, deceased or fictional person.

The proposed criteria would not cover stylized avatars, drawings or caricatures without a realistic likeness. They would also exclude adjustments that improve image or sound quality, along with visual identity elements such as graphics, bumpers and logos.

Nunes Marques also found no electoral violation in an AI-generated video simulating former President Jair Bolsonaro, which appeared at a Liberal Party convention in July. He did not consider the video to be early electoral advertising. In his view, the rules being defined now should apply only to future cases.

“It is at this moment that the Court begins to define, in light of new forms of communication emerging at a rapid pace, the boundaries of its application. In this context, it does not seem possible to infer from the use of artificial intelligence at the party convention an intention to circumvent electoral law,” Nunes Marques said.

The use of deepfakes to favor or damage candidacies is already prohibited, but ministers have differed over how the ban should be interpreted. Resolutions approved by the TSE in March did not prohibit AI in campaign advertising, but required such content to be labeled.

“

It is at this moment that the Court begins to define, in light of new forms of communication emerging at a rapid pace, the boundaries of its application. In this context, it does not seem possible to infer from the use of artificial intelligence at the party convention an intention to circumvent electoral law.