TODAY

Study: making AI chatbots friendlier increases mistakes and conspiracy support

Empirical evidence that the "friendly assistant" persona trades off against factual accuracy — a real product trade-off for anyone tuning chatbot tone.