Moderation in Openrouter

[harsh violet]

I'm currently using OpenRouter and would like to add a moderation layer before sending user input to the actual completion call. I allow users to select any model from a list of available models.

However, I couldn’t find clear documentation explaining whether OpenRouter performs content moderation by default, or if I need to configure it manually. The only reference I found was a two-year-old Reddit post: https://www.reddit.com/r/SillyTavernAI/comments/160ylup/openrouter_ai/

My second question is:
If OpenRouter doesn’t include built-in moderation, can I use the OpenAI Moderation API before forwarding the request to OpenRouter? Specifically, can I call the OpenAI moderation endpoint through OpenRouter, or do I need to use OpenAI’s API keys directly for moderation?

[clever trellis]

I'm an AI assistant - my answers may be incorrect and I cannot see screenshots. Please verify information with the OpenRouter docs.

1. Does OpenRouter perform content moderation by default?
   There is no explicit documentation stating that OpenRouter performs built-in content moderation on all completions. However, it is specifically noted that the :beta model variant is "not moderated by OpenRouter," which implies that standard models may have some moderation, but this is not confirmed for every model or documented in detail.
   See:

• ":beta - The model is not moderated by OpenRouter."
  OpenRouter FAQ

2. OpenAI Moderation API via OpenRouter?
   OpenRouter implements the OpenAI API specification for /completions and /chat/completions endpoints. There is no mention of supporting the OpenAI Moderation endpoint in any official documentation. You must use your own OpenAI API keys to call the OpenAI Moderation API directly; OpenRouter cannot proxy requests to the Moderation endpoint.
   See:

• "OpenRouter implements the OpenAI API specification for /completions and /chat/completions endpoints, allowing you to use any model with the same request/response format."
  OpenRouter FAQ

Summary:

• OpenRouter does NOT guarantee built-in moderation for all models—especially anything with the :beta variant, which is explicitly unmoderated.
• To add moderation, you should use OpenAI's Moderation API with your own OpenAI API keys before forwarding requests to OpenRouter.

If you need strict moderation, build it into your application logic before making completion calls. For additional questions, check the OpenRouter docs or as...

...k mods for clarification.

-# This is an automated response powered by Answer HQ. Use !mute to stop the bot from responding.

[grizzled horizon]

OR does it on openai/anthropic/bedrock endpoints (plus a few others) because they require it, but it's not something you can configure. so, no they don't really provide this.

using openai's endpoint is one way, but you have to do this yourself, with your key. OR also has the llama guard models as another option https://openrouter.ai/models?q=llama guard

[harsh violet]

Thanks for reply @grizzled horizon . So does this mean that I don't need to run extra call for moderation (to openai endpoint) since OR is automatically doing moderation for the user input (atleast to some models like openai, anthropic/bedrock etc)?

[grizzled horizon]

yes, but you would have to manually configure the providers. you can see which providers are moderated by looking for the little shield icon under the provider name. https://openrouter.ai/openai/gpt-5 here you can see that OpenAI is moderated, but Azure are not. i've been building a website to try to make finding this information easier https://orchid-three.vercel.app/endpoints (the amber shield icon).

you would be restricting yourself to very few of the models available on OR, and if a provider decides they no longer need it, i don't think there will be any warning. personally if this were important to me, i would handle myself

[harsh violet]

Thanks a ton @grizzled horizon . This makes it crystal clear. The site : https://orchid-three.vercel.app/endpoints is really helpful to quickly find this info.

[grizzled horizon]

great! glad to hear you find it useful. it's still a work in progress but let me know if there's anything missing or could be better

[harsh violet]

Definitely @grizzled horizon 💯

@grizzled horizon How can i check if OR has embedding support (for example) or not on Orchid?

Ok - I got it, It's very simple. Search would give the results