Brazil banned political deepfakes. 60% of AI-generated
Researchers identified 554 deepfake posts impersonating presidential candidates and journalists during the 2026 Brazilian election in violation of electoral.
- Reported date
- Oct 1, 2026
- Target
- Luiz Inácio Lula da Silva and Flávio Bolsonaro
The exact incident date was not established. This entry is dated by its source report.
During the 2026 Brazilian presidential election, researchers identified hundreds of instances where deepfake technology was used to impersonate political figures and journalists, despite strict national prohibitions. The findings highlight significant challenges in enforcing electoral regulations regarding synthetic media in a polarized political environment.
What happened
Project VigIA, a collaborative research initiative, monitored political content between August 16 and September 28, 2026. Out of 920 confirmed pieces of AI-generated or manipulated content, 554 were classified as deepfakes—media designed to make real individuals appear to say or do things they never did.
These deepfakes frequently targeted the two leading presidential candidates: President Luiz Inácio Lula da Silva and Senator Flávio Bolsonaro. Lula was the most frequently depicted politician, appearing in 379 AI-generated posts, while Bolsonaro appeared in 190. The content was used both to promote and attack the candidates. Beyond the candidates, the deepfakes also impersonated well-known Brazilian television journalists, creating fabricated reports regarding election polls and results. Other instances involved the unauthorized use of celebrities to falsely endorse political figures.
Although Brazilian electoral law mandated the disclosure of synthetically created or altered material and completely prohibited deepfakes in official advertising, compliance was low. Nearly 60% of the confirmed AI-generated posts lacked any disclosure. Enforcement proved difficult, as 47% of the content originated from individual social media users or activists, and 27% came from anonymous or avatar-based accounts. Candidates themselves were responsible for 24.5% of the confirmed AI content, directly violating the established legislation.