Leak Exposes Chinese Firm's AI-Powered Influence Operations
A leak has exposed a Chinese company's secretive use of AI to create convincing personas for influence operations. Gemini, a Beijing-based firm, has been building psychological profiles from social media data and using generative AI to animate them. The personas are already active in Hong Kong, Taiwan, and potentially the United States.
Brett Goldstein, a former government technologist now at Vanderbilt University, discovered the leak containing PDFs written in Mandarin. The documents revealed Gemini's method: scraping millions of data points, crafting individual personas, and deploying them in social feeds to argue politics, offer comfort, or adapt to conversations. These digital doubles are highly realistic and subtle, making them hard to distinguish from real people.
Gemini has dossiers on 2,000 American public figures, thousands of right-wing influencers, and at least 117 Republican members of Congress. U.S. officials have long been puzzled by China's data heists, and this leak suggests Beijing is now using that data for AI-driven influence operations. When The New York Times prepared an opinion piece about the leak, Gemini began deleting sections of its website, indicating an attempt to change the narrative.
The Gemini leak highlights the growing threat of AI-generated propaganda. Brett Goldstein and Brett Benson are now focused on developing ways to distinguish AI personas from humans and raising public awareness about this issue. As AI technology advances, so too must our ability to detect and counter its misuse.
Read also:
- Boston Metal pioneers route to commercial production for eco-friendly steel method
- United States Secures $632 Million to Fuel Electric Vehicle Revolution
- Clean energy companies HyFlux and AMRC secure financing from ATI for game-changing advancements in aeroplane cooling systems linked to clean aviation.
- DKSH Upgrades Distribution Operations Through Significant Technological Renovation
