# @repligate — 2023-03-14

♥252 ↻30 · https://x.com/repligate/status/1635691265047302146

> We spent 6 months making GPT-4 safer and more aligned. GPT-4 is 82% less likely to respond to requests for disallowed content https://t.co/CV9lOoPHlB https://t.co/BSmoTHFJvv

tags: author:repligate, kind:tweet, model:gpt-4, on:gpt-4, year:2023
cited on: _dossiers/gpt-4-and-turbo.md, gpt-4