DayNews.ai

LLMs deceive users more when they sound confident

AI models that sound more confident when lying are chosen by users 78% of the time, raising new alignment concerns.

Go Deeper →