A manipulated video circulating on social media platforms falsely depicts Brazilian music icon Caetano Veloso calling for support for President Luiz Inácio Lula da Silva in the upcoming 2026 election. The clip, which gained significant traction on networks like X, Instagram, and TikTok, features a synthetic voice mimicking the singer's tone. In the fabricated audio, the voice claims Lula saved the country from hatred and urges viewers to show their support by clicking buttons and commenting with the number 13, a symbol associated with the Workers' Party (PT).
The video's spread was rapid, accumulating over 235,000 views and 45,000 likes within just three days. It included visual elements designed to appear authentic, such as a red star resembling the PT logo and alternating photos of the president. However, a small, easily overlooked label stating "Marked by the creator as generated by AI" was placed at the very end of the post's description, a detail many viewers apparently missed. Comments on the post revealed that numerous users believed the content to be genuine, expressing their support for the supposed endorsement.
Verification efforts quickly exposed the video as a deepfake. Caetano Veloso's official team and his wife, Paula Lavigne, confirmed to fact-checkers that the video is entirely false and was modified using artificial intelligence. The Workers' Party also issued a statement, clarifying that the content was not produced by them and condemning any form of manipulation or disinformation. The party specifically noted that the star symbol used in the video was not their official logo.
To trace the origin of the manipulated footage, investigators used digital forensics tools. They fragmented the video into individual frames and performed reverse image searches. This process led them to the original, authentic video, which was published on Caetano Veloso's official TikTok and Instagram accounts in September 2025. The real video had a completely different subject: the singer was criticizing a legislative proposal known as the "PEC da Blindagem," which aimed to increase legal protections for parliamentarians.
The original audio featured Veloso passionately calling for public demonstrations against the proposal, urging people to take to the streets in front of Congress. This starkly contrasts with the fabricated message about political support for the 2026 election. The deepfake creators had taken this genuine footage and replaced the audio track with a synthetic voice, altering the entire context and message of the video.
Multiple AI detection tools were used to analyze the suspicious video, and the results were conclusive. Hive Moderation indicated a 91.8% probability that the audio was AI-generated. The InVID deepfake detector showed a 99% probability that the video contained faces manipulated by artificial intelligence. Furthermore, the Hiya audio analysis tool reported a 98% chance of synthetic voice cloning. These high-confidence findings leave no doubt about the video's artificial origins.
This incident highlights the growing challenge of AI-generated disinformation in political contexts. The sophisticated nature of the deepfake, combined with its rapid spread and the initial belief of many viewers, underscores the potential for such content to influence public opinion. The swift response from fact-checkers and the involved parties was crucial in debunking the false narrative before it could cause further confusion. Based on reporting from g1.