This sounds like damned if you do, damned if you don't for OpenAI. If OpenAI doesn't implement these "filters" in results, then people would have been up in arms same way. Facebook and Twitter are already complained about a lot for allowing hate speech similarly.
Blanket censorship is saying I can’t have steak because a baby might choke on it.
The Azure AI service enforces a separate profanity filter on top of the base ChatGPT model. That’s not even OpenAI, it’s Microsoft filtering the output with regexes!
OpenAI themselves are not forced to release exactly one model. After the base training, they could have fine tuned it for different levels of Puritanism, including none for people that want a tool, not a priest.
