Users report GPT-6 Astra tightens safety refusals, becoming more Anthropic-like

TheStalwart · x · 2026-09-15

TheStalwart notes GPT-6 Astra is going "next levels" with personal safety, sharing refusal screenshots.

In replies, ivanbezdomny adds that GPT was previously very permissive with almost no refusals, but since Astra it has become noticeably more Anthropic-like in its guardrails — though "a bit dumber" at it since OpenAI is newer to this style of moderation. The thread captures community perception that the new model ships with much more conservative safety behavior.

Related event: Users report GPT-6 Astra refuses more, resembling Claude's style(2 posts)→

Original post →

More from Models

Models channel →