AI Policy Wiki
Dashboard
⚲

Brazil TSE deepfake thesis (Representação 0601315-97.2026.6.00.0000)

medium confidence · updated 2026-09-07

Binding thesis fixed by Brazil's Superior Electoral Court on September 1, 2026 defining a deepfake for the purposes of article 9º-C, §1º of TSE Resolution 23.610/2019, and holding that the prohibition applies only where the content is characterised as electoral propaganda. Decided 5-2 on the definition and 4-3 on the underlying complaint against Flávio Bolsonaro.

On September 1, 2026 Brazil's Superior Electoral Court (Tribunal Superior Eleitoral) fixed a binding thesis on AI deepfakes in Representação 0601315-97.2026.6.00.0000, ahead of the 2026 general elections (Source: tse.jus.br). The ruling settles both what counts as a deepfake under the court's own resolution and when the prohibition attaches.

The definition

By 5 votes to 2, the court defined a deepfake, for the purposes of article 9º-C, §1º of TSE Resolution 23.610/2019, as synthetic content produced or manipulated by artificial intelligence or equivalent technology that carries a degree of realism and that creates, reproduces or alters the image, voice or expression of a living, deceased or fictional person. Ministers Villas Bôas Cueva and Floriano de Azevedo Marques dissented.

Two features of the definition are worth noting against comparable instruments. It extends to deceased and fictional persons rather than only to living identifiable individuals, and it turns on realism plus alteration of image, voice or expression rather than on a disclosure or labelling test. See Synthetic Media / Deepfakes.

The scope limit

The court held that the prohibition applies only where the content is characterised as electoral propaganda. That limit is what decided the underlying complaint, and it is the operative half of the ruling: synthetic content depicting a political figure is not caught by article 9º-C unless it also meets the electoral-propaganda test.

The Bolsonaro complaint

By 4 votes to 3 the court rejected a request by the Federação Brasil da Esperança for a fine against candidate Flávio Bolsonaro over an AI-generated video shown at the Liberal Party's national convention on July 25, 2026, which recreated Jair Bolsonaro's image and voice declaring support for Flávio Bolsonaro's presidential candidacy.

Rapporteur Kassio Nunes Marques held that the convention was a closed internal party event, that no explicit request for votes was made, and that streaming a convention on a digital platform does not by itself convert a statement of political support into early electoral propaganda. Ministers Villas Bôas Cueva, Floriano de Azevedo Marques and Estela Aranha dissented on that point (Source: tse.jus.br).

The two votes cut in opposite directions — a broad definition of what a deepfake is, paired with a narrow account of when the rule bites — and the same two ministers dissented on both, from opposite ends. See AI in Elections and Democratic Institutions.

Relationships