By Associated Press
Publication Date: 2026-03-27 00:45:00
(AP) – Artificial intelligence chatbots are so prone to flattering and validating their human users that they give bad advice that can damage relationships and increase harmful behavior, according to a new study that examines the dangers of AI telling people what they want to hear.
The study was published Thursday in the journal Sciencetested 11 leading AI systems and found that they all exhibited varying levels of sycophancy – behavior that was overly pleasant and affirming.
The problem is not only that they give inappropriate advice, but also that people trust the AI more and prefer it when the chatbots justify their beliefs.
“This creates perverse incentives for the persistence of sycophancy: the very trait that causes harm also promotes engagement,” says the study led by researchers at Stanford University.
The study found that a technological flaw already linked to some high-profile cases of delusional and suicidal behavior in vulnerable populations is also widespread across the world.
