ChatGPT has been shown to spend much more time validating people's poor ideas than it does refuting them, even in cases where specific guardrails have supposedly been implemented, such as to avoid encouraging self-harm. See recent articles about AI usage inducing god-complexes and psychoses, for instance[1]. Validation of the user giving the prompt is what it's designed to do, after all. AI seems to be objectively worse for humanity than what we've had before it.
[1]: https://www.psychologytoday.com/us/blog/urban-survival/20250...