The deepfake threat just got a little more personal | CSO Online

Researchers at Google DeepMind and Stanford University developed AI replicas of over 1,000 individuals using two-hour interviews. These AI models, trained on generative AI, demonstrated 85% accuracy in matching human responses on personality tests and social surveys, raising concerns about potential misuse for deepfake attacks. The researchers propose an “agent bank” for controlled research access to these simulated personalities.