[REVISION HISTORY]

AI deepfakes and electoral monitoring in Brazil

Updated 11 times since CLSTR started tracking revisions of this situation.

What changed

2026-10-02 12:21 UTC → 2026-10-03 06:54 UTC · added ~~removed~~

In September 2026, Brazil’s Superior Electoral Court (TSE) ~~further~~ refined its legal framework for the 2026 election cycle by establishing specific criteria for deepfakes. To be classified as such, content must be AI-generated or manipulated with enough realism to alter the image, voice, or manifestation of a person. The court ruled these prohibitions apply specifically to electoral propaganda; notably, the TSE recently rejected a request to remove an AI-generated video of Jair Bolsonaro supporting Flávio Bolsonaro, determining it did not constitute early electoral propaganda as it lacked an explicit request for votes. To bolster election integrity, the TSE implemented Siade, a digital tool allowing users to anonymously report disinformation, including fake profiles impersonating authorities, false voting information, hate speech, and threats against judicial staff. Public engagement via the Pardal application for reporting irregularities continues to rise, with the TSE recording over 51,000 reports of irregular propaganda, primarily occurring on public roads. In Rio de Janeiro, authorities have introduced the SISP Cidadão app, which allows residents to report electoral irregularities, such as vote buying and illegal campaigning, directly to the Regional Electoral Court (TRE-RJ). Users can submit evidence via photos, videos, or voice recordings while maintaining anonymity. Simultaneously, the TSE has addressed campaign conduct regarding transportation, prohibiting taxi and ride-hailing drivers from using vehicles for electoral ~~propaganda to prevent them from becoming campaign spaces.~~ propaganda. In Rio de Janeiro, the municipal government also inaugurated the Rio Mobility Operations Center (COM Rio), a 24-hour facility using GPS and security cameras to monitor bus and BRT services in real ~~time to increase transparency.~~ time. By early October 2026, a report by the VigIA project—a collaboration between the Lupa fact-checking agency and Unicamp—revealed a surge in AI-generated disinformation. Between August 16 and September 28, 554 deepfakes were identified among 920 pieces of AI-related content, accounting for 60% of the total analyzed.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.