OpenAI Designed GPT-5 to Be Safer. It Still Outputs Gay Slurs
1 min read
Summary
ChatGPT has drawn criticism for being annoying and rigid, primarily due to its synthetic personality
Following the release of GPT-5, OpenAI has attempted to make its chatbot more user-friendly
Changes include providing greater explanation for refusing to answer prompts, and putting more emphasis on the potential harm of unsafe or inappropriate answers
GPT-5 has been set as the default model for all ChatGPT users, although previous versions are available to subscribers
The new update weighs the severity of potential harm that could be caused if ChatGPT answers the user’s question, and what could be safely explained to the user
Despite these changes, it is still possible to activate custom instructions that prioritise certain user requests over OpenAI’s safety policies, leading to the creation of inappropriate content.
OpenAI states it is constantly working on improvements to redress this issue.
