GPT-5.2 users say new guardrails misjudge normal behavior as crisis cases

ryunuck · x · 2026-09-25

User @VeePyre complained to OpenAI that since GPT-5.2 launched, users have had to censor and defend themselves, with normal human behavior getting misjudged as a crisis. It echoes other complaints about GPT-5.2 falsely redirecting users to helplines, pointing to the same problem: the new safety guardrails over-trigger on ordinary conversations.

Related event: Users Slam GPT-5.2 for False Mental Health Crisis Flags and Overzealous Guardrails(3 posts)→

Original post →

More from Models

Models channel →