Brazil defines what an electoral deepfake is
The ban applies only when synthetic content is electoral propaganda.

On September 1, Brazil’s Superior Electoral Court decided a case that became a benchmark for AI use in the 2026 campaigns. The Brazil of Hope Federation (PT, PCdoB and PV) filed a complaint against Flávio Bolsonaro over an AI-made video that recreated Jair Bolsonaro’s image and voice endorsing his son’s presidential bid, shown at the PL’s national convention on July 25. It sought removal of the video and a fine against the candidate.
By 4 votes to 3, the court rejected the complaint. The rapporteur, Kassio Nunes Marques, argued that the video was shown at a closed party event aimed at delegates and leaders, and contained no explicit request for votes; it was therefore not early electoral propaganda. Villas Bôas Cueva, Floriano de Azevedo Marques and Estela Aranha dissented.
More important than the case was the thesis the full court set, by 5 votes to 2, to guide electoral judges. Under Article 9-C of Resolution 23.610/2019, a deepfake is synthetic content produced or altered with AI or equivalent technology, with a degree of realism, that creates, reproduces or alters the image, voice or statements of a living, deceased or fictional person. And the deepfake ban applies only when that content qualifies as electoral propaganda.
The second part of the thesis addresses digital broadcasting: streaming a convention online does not by itself turn political support aimed at delegates into early propaganda. The court called for each case to be judged by its content, language, audience and context, not just its online reach.
The TSE paired the rule with tools. On August 26, together with Google, it launched YouTube’s likeness-detection feature for candidates: after verifying their identity with an ID and a selfie, candidates receive alerts when videos reproducing their image appear and can request removal under the platform’s privacy policies. Google Brazil’s president stressed that the protection matters especially for women, frequent targets of image-based attacks.
For consultants, the conclusion is that AI’s legal limit lies not in the tool but in the context: the same piece may be legitimate at an internal event and punishable if circulated as propaganda. Every synthetic asset needs legal review before release, and every campaign should enroll its candidate in the available detection systems.
The legal line on AI is not the tool but the context: every synthetic asset needs legal review.
Summary written by the FOCUS editors based on the original story. Rights to the original content belong to its publisher.
FOCUS courtesy space
NewsThe news, when it matters.News, analysis and election coverage on every screen.nmas.com.mx →

