AI-Generated Deepfake Ads on Instagram Defraud Millions in Brazil
Brazilian police arrested four suspects and froze assets in October 2025 in a large-scale fraud scheme that used AI-generated deepfake videos of celebrities including Gisele Bündchen to advertise non-existent products on Instagram and Facebook, defrauding millions of dollars from victims. The criminal network was led by a person who also operated a digital crime instruction service via Instagram.
What happened
Brazilian police launched "Operação Modo Selva" in October 2025 against a criminal network that used artificial intelligence to generate deepfake videos of supermodel Gisele Bündchen and other celebrities including Angélica Huck, Juliette, Maísa and Sabrina Sato to promote non-existent products through fraudulent Instagram and Facebook ads.
The scheme featured deepfake videos falsely showing the celebrities endorsing products such as a fictitious anti-wrinkle kit and fake suitcase giveaways, which directed victims to fraudulent websites to pay shipping fees for items that never arrived. The investigation revealed a hierarchical criminal organization led by Levi Andrade da Silva Luz, who also operated a "digital crime university" via an Instagram profile teaching others to become "digital predators."
Reported impact
The fraud scheme moved over 20 million Brazilian reals (approximately $3.9 million USD) through suspicious funds identified by federal anti-money laundering authorities. Most victims lost small amounts, typically under 100 reals (approximately $19 USD), and many did not report the crimes.
Index analysis
The operation resulted in the arrest of four suspects, with three remaining at large. Police issued 26 judicial warrants across five states (Rio Grande do Sul, Santa Catarina, Pernambuco, São Paulo and Bahia) and froze assets potentially totaling up to 210 million reals.
Governance controls
Meta stated that its policies prohibit "ads that deceptively use public figures to try to scam people" and said it removes such ads "when detected," and that it employs "specialized systems to detect celeb-bait" and trained review teams. Brazil's Supreme Court ruled in June 2025 that social media platforms can be held liable for criminal ads posted by users if they fail to act swiftly to remove the content, even without a court order.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.