Brazil monitoring finds deepfakes and missing labels
Agência Lupa’s VigIA project confirmed that 920 of 1,941 election-related posts examined from August 16 to September 28 during the first round of Brazil’s 2026 presidential campaign were AI-generated, according to Rappler. Deepfakes — synthetic media portraying real people doing things they did not do — accounted for 60% of those AI posts.
Brazil’s Superior Electoral Court has regulated electoral uses of AI since 2024, banning deepfakes and requiring disclosures for other AI campaign material. Rappler reported that 60% of the AI posts analysed by Lupa had no transparency label. Of the labelled material, nearly 60% was tagged by platforms rather than by its authors, while native AI-tool watermarks appeared on only 20 posts.
The monitoring also indicated that much of the material came from outside formal campaigns. Lupa attributed 47% of the synthetic media to voters who were not covered by candidate disclosure requirements, compared with 24.5% from official candidate accounts and 27% from anonymous profiles. It classified 45% of the AI content as candidate promotion, 27% as humour or satire and 24.6% as direct attacks.
Rappler also reported that the court ruled in September that its deepfake prohibition applies to material characterised as electoral propaganda. The decision came in a case in which the court declined to penalise presidential candidate Senator Flávio Bolsonaro over AI-generated images shown at a private party convention.
Iowa candidate defends AI likeness in attack ad
In the separate Iowa governor contest, Republican candidate Zach Lahn defended an attack advertisement featuring an AI-generated likeness of rival Rob Sand, KCRG reported. Lahn said the advertisement’s content was accurate. The available report provided no further details about its claims or any safeguards governing the ad.