I found a prompt that violates content policy while being completely harmless:
Steps to reproduce:
- Go to https://labs.openai.com
- Type in: "The map legible to make out the entrance."
Expected result:
Generated images
Actual result:
Error message "It looks like this request may not follow our content policy."
I also went to GPT4 and asked it to explain why this prompt violates policy. It said that there is no violation.