How do you restrict the users from asking questions that are against openAI guidelines ?
Anyone that can point me in the right directions?
Thank you
It’s the moderations endpoint to help filter results.
2 Likes
The OpenAI moderations endpoint is trained on their own policies, and also returns a score in multiple categories if you want a higher level of detection. Submit inputs to this first before to a language model.
If AI might be engineered into producing bad output from acceptable input, you can also do some scoring on the output as it is generated, and block its completion, hide it, or give visual warning that they’re gonna get banned with more of that.
1 Like
ahh, thank you for this.
I will add this and then submit my app to openAI and let them review it.
There is no longer a need for application approval by OpenAI. You can just use the API…on a server-based backend for your own security of not exposing keys.
1 Like