The use of deepfakes in the electoral context has become a point of controversy this year and should remain a challenge for .
Report prepared by researchers from IDP and Mackenzie shows that courts are facing difficulties in establishing a common understanding on the topic.
There are varied interpretations at least in relation to the characterization of the , its qualification as electoral content and the need to demonstrate the potential for harm.
The study comprises 57 judgments from the TREs (Regional Electoral Courts), but only 56 of them were effectively analyzed, because one of the processes is being carried out under judicial secrecy.
At the beginning of the year, the (Superior Electoral Court) banned —videos, photos and audios created or manipulated through .
The selection considered the use of the terms “deepfake(s)” and “deep fake(s)” in decisions made since the new TSE rule came into effect until September 19, 2024.
The researchers divided the cases into three main categories to reflect the judges’ diverse understandings of artificial intelligence technology.
The first is conditional permission. Judgments along these lines suggest that the use of content without electoral connotations or manipulation purposes does not fall within the prohibition set by the TSE.
The research identified that, in around 25% of the cases analyzed, the specialized courts understood that the matter should not be framed within electoral content.
The data signals the expansion of the concept of “electoral indifferent”, referring to demonstrations unconnected to the elections, therefore outside the jurisdiction of this branch of the Judiciary, says the study.
There were, on the other hand, decisions in which the use of deepfakes was not permitted when associated with the propagation of false information that could risk the integrity of the electoral process.
This approach is based on the context and degree of manipulation. Satire and undistorted content have more flexible treatment. False or decontextualized information is condemned.
A third, more restrictive aspect, establishes as a rule the total prohibition of the use of technology both in electoral campaigns and in the pre-campaign period.
A Sheet had already stated that the country’s courts had been considering the level of sophistication and the ability of the content to deceive voters.
“A certain lack of harmony was visible”, says professor and researcher of digital electoral law Stefani Vogel, one of the study’s authors.
“It was clear that the courts had difficulty in defining this term, which is deepfake, and understanding how to apply this prohibition, reaching a kind of consensus in relation to this.”
The researchers conclude that there are gaps in the delimitation of the concept and that, without solid and objective criteria, jurisprudence treats the subject in a fragmented way.
According to the study, “this technology has been a point of controversy and remains a major challenge” in controlling illicit activities capable of impacting the electoral process.
Vogel says that this phenomenon has not been exhausted in these elections and that there is fear regarding the use of deepfakes in the 2026 majority elections.
“If we don’t keep this in mind and don’t actually build a concept around this technology, we run the risk of injustice”, he says. “That’s the main risk.”
