A Deepfake Jensen Huang Livestream Fooled Over 100,000 Viewers in a Crypto Scam In a stark demonstration of how convincing AI-generated media has become, a fake livestream featuring a deepfake of NVIDIA CEO Jensen Huang attracted over 100,000 simultaneous viewers. The fraudulent broadcast ran concurrently with NVIDIA’s actual GPU Technology Conference keynote, dramatically overshadowing the real event which only had about 20,000 live viewers. The AI-generated version of Huang was used to promote a cryptocurrency scam. The deepfake spoke about a supposed crypto mass adoption event that aligned with NVIDIA’s mission, urging viewers to scan a QR code and send in cryptocurrencies. The premise of sending money to one of the world’s most valuable companies to advance human progress was, on its face, an obvious fraud, though it is unclear if any of the viewers actually fell for the financial pitch. The fake video has since been removed from the platform. Before concluding that this event signals a complete breakdown of reality, some important context is needed. Little is known about the 100,000 accounts that tuned into the fake stream. It is uncertain if they were all real people or where they were located. The fake stream’s high viewership could also be attributed to aggressive promotional tactics, and it is unknown how long the average viewer stayed after clicking. For comparison, the official NVIDIA keynote has since accrued over 200,000 total views after the live broadcast concluded. The success of the deepfake is also a function of its subject. Jensen Huang has hosted numerous GTC conferences over the years, resulting in a vast library of public footage of him speaking on stage. This provides ample training data for malicious actors to create a convincing replica. Furthermore, the actual NVIDIA event may not have captured mainstream attention as it focused on announcements like a partnership with Uber for autonomous driving and government contracts, rather than new consumer graphics cards. This incident is part of a growing trend of hyper-realistic AI media being used for deception and parody. Just earlier this month, OpenAI CEO Sam Altman donated his likeness to the Sora 2 AI video generator. Users almost immediately created videos of Altman engaging in outlandish acts, from shoplifting GPUs to eating a grilled Pikachu. While this specific deepfake stream may not mean society is completely cooked, it is a clear sign that the water is getting dangerously hot. The ability to clone a prominent CEO and draw a massive audience in real-time marks a significant escalation in the potential for AI-powered disinformation and scams. The article originally appeared on Engadget.
