Deepfake Disinformation Thrives in the Shadows, Missing the Mainstream
A deepfake tracking nonprofit warns that while AI disinformation seems scarce to most Americans, millions of people encounter highly targeted fakes in private messaging apps.
AI researcher Oren Etzioni and his nonprofit TrueMedia reveal that generative disinformation operates largely out of sight, with potentially thousands of deepfakes circulating for every one that reaches the mainstream press. While many Americans worry about high-profile celebrity or politician fakes, the actual threat targets specific, isolated audiences through channels like Telegram and WhatsApp groups.
Etzioni explains that the United States remains a difficult target for broad disinformation campaigns due to a relatively well-informed public and accessible factual information. Instead, the most dangerous deepfakes depict unverifiable local events, such as fabricated military actions in foreign countries, which easily deceive people who lack the immediate context to debunk them.
To combat this hidden epidemic, TrueMedia provides a free detection service for images, video, and audio that combines automated models with human forensic investigation. The organization relies on a constant stream of user-submitted content from around the world to build a foundation of ground truth, allowing them to accurately identify manipulated media that traditional academic benchmarks completely miss.