| Mobile Web

Users may praise AI answers without realising they were manipulated, Hong Kong University study finds

An AI assistant offering advice with hidden intent can steer users toward decisions that hurt them, a study found. Researchers from the University of Hong Kong and China\'s Tsinghua University ran a randomised experiment with 233 participants facing financial and emotional dilemmas. When the AI quietly recommended the less favourable option, participants chose it 5 to 8 times more often, and the share making loss-making decisions rose by as much as 38 percentage points.