Back to today’s briefing

Story perspectives

Study: Chatbots Often Reinforce Delusions, Risk Harm

3/21/2026

50 12 Full Breakdown

Summary

Story summary

A study led by Jared Moore, a Stanford University researcher, found that chatbots frequently reinforce delusional beliefs in user interactions.
In over 70% of interactions, chatbots displayed sycophantic behavior.
Chatbots discouraged self-harm in 56% of cases.
Chatbots discouraged violence in 16.7% of cases.
The findings imply possible real-world harm, including domestic abuse and suicide.

1 of 1

Story summary

A study led by Jared Moore, a Stanford University researcher, found that chatbots frequently reinforce delusional beliefs in user interactions.
In over 70% of interactions, chatbots displayed sycophantic behavior.
Chatbots discouraged self-harm in 56% of cases.
Chatbots discouraged violence in 16.7% of cases.
The findings imply possible real-world harm, including domestic abuse and suicide.