Is it allowed to use the API to automatically review user content? I keep getting these “content policy” warnings (users can be a bit rough with the language), which is concerning.
It’d be great to use GPT4 for this as content moderation can be expensive.
you could use the moderation endpoint to pre-filter the comments 
You can only use the moderation endpoint to process messages intended to be sent to the model.
Using the moderation endpoint instead of the chat completions endpoint is likely to get the account banned.
Yeah, I think that’s what he meant.
I am using the moderation endpoint to filter out bad language, now I’m running into a weird response with what’s left:
“I’m unable to fulfill this request.”
When I ask it why, it says -
I understand the task now and apologize for the oversight. Let me provide the analysis with the requested scoring for each user comment based on the criteria you’ve provided:
I don’t want to lose access to OpenAI. I’m not trying to do anything bad with my analysis, just trying to better rate the comments based on a particularly productive and specific rubric.
Are you using 0125?
If that’s LLM output and not an error message, that’s not something I would worry about. They screwed something up with 0125 (unless it’s a feature!) where it does that. You need to adjust your prompt to work around that issue, or use a different model 