Discourse has expanded access to its AI features, now available on Pro and Business plans in addition to Enterprise. Users can enable AI features via the [Discourse AI guide](https://meta.discourse.org/t/disco urse-ai/259214), though LLM-based features require an API key from a supported provider. A user (roniw) raised privacy concerns, noting that even after configuring a custom LLM provider, Discourse-hosted LLMs remain visible and cannot be removed — a concern for EU-hosted sites with strict data governance. Jagster clarified that CDCK (Discourse’s hosting) complies with EU data regulations. sam confirmed users are never forced to use Discourse’s LLMs, and while built-in models can’t be deleted, a “hide” feature may be added. Discourse self-hosts its LLMs and does not train on user data. Saif reiterated that users can opt entirely into their own LLM providers and provided a link to the LLM settings page for configuration details. Privacy-conscious users on Business plans can safely use their own LLMs without data exposure to Discourse’s models.
We are excited to announce that in addition to Enterprise, all Discourse AI features are now available on our Pro and Business plans.
Start here to read more about Discourse AI and how to enable the plugin and individual features for your site.
Note that Large Language Model (LLM) based features require an API key for a supported provider
We hope this opens the doors to more people using Discourse AI.
2 posts were split to a new topic: Enabling Discourse AI on Pro Hosting Plan
I would like to try these new features and I can’t find information about privacy.
I configured our own LLM provider, but I can’t remove the CDCK Hosted Large LLM and other Discourse default providers. We don’t want our data accessed by a model without a specific privacy agreement. We are using the Business plan and are hosted in the EU 1 datacenter. What options do we have?
Here is two different things now:
First one is easy. I’m sure CDCK follows regulation, and in this matter they are more acting as a middle man and the last point defenetly follow rules.
Second one… I’m sure someone from CDCK will answer you, and they have support channels (is this one of those, I don’t know).
The good news is that we will not use any of our llms unless you opt a feature to using it.
We do not support deleting built in llms, but I think @Falco we could add a “hide” feature for cases like this
We do self host the discourse llm models and do not share any data with third parties
Thank you for the information. It would be very helpful to be able to opt out of your llms and just use our provider. Do your llm models train on our data if we turn them on?
You are not obligated to use our Discourse-hosted LLMs to power the AI features. We use open-source models and do not train on your data.
Discourse AI is flexible and supports various numbers of providers, please check out Discourse AI - Large Language Model (LLM) settings page and each of the AI features you are interested in for more info.
