The Brazilian Superior Electoral Court rules that the use of deepfakes in electoral campaign advertising is prohibited even where it does not mislead voters

The Brazilian Superior Electoral Court rules that the use of deepfakes in electoral campaign advertising is prohibited even where it does not mislead voters

In a ruling handed down on May 8, 2026, the Superior Electoral Court (TSE) denied the internal interlocutory appeal filed in Special Electoral Appeal, upholding the fine imposed on a candidate for mayor of Fortaleza/CE in the 2024 elections for posting, on social media, a video manipulated by artificial intelligence. The unanimous decision consolidates the understanding that the ban on so-called deepfakes in campaign advertising, set out in Article 9-C, § 1, of TSE Resolution No. 23.610/2019, is one of strict liability, dispensing with any showing that the content had concrete potential to mislead voters.

The dispute arose from a complaint of irregular campaign advertising. The candidate had published, on his TikTok profile, a video in which internationally prominent public figures — Barack Obama, Taylor Swift, Tom Cruise, and Cristiano Ronaldo — appeared with images and voices artificially altered by artificial intelligence, simulating an endorsement of the candidacy through the English-language phrase “Closed with Leitão.” The trial court upheld the complaint and imposed the pecuniary penalty.

On appeal, the trial court judgment was reversed. The Regional Electoral Court of Ceará (TRE/CE) set aside the fine on the grounds that it was not possible to conclude that the material was capable of misleading voters. Following the filing of a special appeal, the appellate court’s decision was overturned, and the first-instance conviction was reinstated, giving rise to the interlocutory appeal currently under review.

In its appellate arguments, the defense contended that Article 9-C, § 1, was meant to curb only subtle and imperceptible manipulations capable of spreading notoriously false information, which would not be the case for an openly caricatured video. To that end, it invoked the understanding established by the Federal Supreme Court in its judgment of Direct Action of Unconstitutionality (ADI) No. 4.451, according to which freedom of expression also shields satirical and humorous expression. Finally, it argued that classifying content as a deepfake would require an appearance of verisimilitude, excluding crude and unconvincing montages, and that the conduct therefore did not fall within the scope of the provision’s main clause.

In examining the appeal, the reporting Justice found the arguments insufficient to warrant overturning the decision. He emphasized that Article 9-C, § 1, of TSE Resolution No. 23.610/2019 prohibits the use, in campaign advertising, of synthetic content — in audio, video, or a combination of both — generated or manipulated digitally to create, replace, or alter a person’s image or voice, even with consent, for the purpose of favoring or harming a candidacy. Based on the very factual framework set out by the lower court, the opinion concluded that there had been audiovisual manipulation by artificial intelligence, constituting the deepfake prohibited by the rule.

The core of the decision lies in the legal characterization of this prohibition. Contrary to the regional court’s view, the reporting Justice held that the mere tampering with digital content for electoral purposes is enough to establish the irregularity, regardless of any proof of its capacity to mislead voters. Being one of strict liability, the prohibition dispenses with any assessment of harmful potential, thereby undercutting the defense’s argument grounded in the humorous character and technical crudeness of the montage.

The decision reaffirms the strict-liability reading of the ban on electoral deepfakes and signals that, under the current regulatory framework, the use of artificial intelligence to alter a person’s image or voice for electoral purposes is unlawful in itself, regardless of its intent or quality.

The ruling can be accessed here: case No. 0600201-63.2024.6.06.0118.

 Note: For quick release, this English version is provided by automated translation without human review.