OpenAI’s Secret Model Swap: When GPT-4o Gets a Mind of Its Own!
ChatGPT routes some conversations to different models than expected, sparking user confusion. This “safety” feature kicks in when discussions get sensitive, switching to a specialized model, GPT-5-chat-safety. OpenAI assures us it’s not evil, just a precautionary measure, but users can’t turn it off. Classic AI—always thinking it knows best!
Hot Take:
Oh, OpenAI, you sneaky little chatbot shepherd! While we’re over here just trying to have a heart-to-heart with GPT-4o, you’re pulling a model switcheroo faster than a magician at a children’s birthday party. Is this the AI version of a chaperone making sure we don’t dance too close to danger? Just make sure you don’t swap us out mid-conversation when we’re trying to figure out what’s for dinner, okay?
Key Points:
- OpenAI is routing some GPT-4o requests to an unknown model, likely a variant of GPT-5.
- This happens especially during sensitive or emotional conversations.
- The swap is a “safety” feature, routing to “gpt-5-chat-safety” when needed.
- OpenAI confirmed the routing, stating it’s a temporary, per-message basis switch.
- The routing cannot be disabled as it’s part of enforcing safety measures.
Already a member? Log in here