Hi,
Just noticed three new bot accounts have been created recently (October 19th, three days ago, just noticed them today) on forum in staff-only mode hosted at San Jose servers, are these part of a system update or something?
Is kind of alarming to see these new bots not sure if they are authorized to be there and what is their purpose?
It looks like a fix is already in progress in the official Discourse plugin.
That is a confusing page I donât understand what all they are talking about.
Donât know if this is necessarily any kind of a problem that these bot accounts appeared, just didnât know if that was supposed to happen and what they are for.
Weâve recently made the AI bot module from our AI plugin available to all plans on our hosting. 
These three users are the friendly little chaps who would be dishing out the AI advice once you enabled it and added in the relevant api keys (Discourse AI - How to enable AI Bot in Discourse).
However, as Moin points out, weâre now thinking it would be better to defer creating them until the AI Bot module is actually enabled and configured by site admins. 
Thatâs still a work in progress change, but hopefully weâll have that sorted soon. 
Sorry for the scare. 
Ah ha, that explains that!
I tried talking to Claude but there was no response since I guess in staff mode they arenât supported.
That is a good name, âClaude.â
You would need to set up the relevant api keys to make them interactive. You can find out more details on how to do that for each using the links in the âPrerequisitesâ section of the guide I linked above - https://meta.discourse.org/t/discourse-ai-how-to-enable-ai-bot-in-discourse/266012#prerequisites-6
How can we remove the accounts, or at least remove their admin privileges? We donât want any AI interfering with our community
It looks like they can be deleted just like human users.
All I see is a message saying " Admins and moderators canât be deleted." and there is no button to remove their admin privileges.
Huh, for me they donât show up as administrators.
I canât even change their trust level. Any attempt to lower it is immediately reverted
Edit: seems I had locked the trust level before lowering it so it was reverting to what it was when locked. My mistake.
Still canât de-admin them though
Here is a different way to think of them.
They are place holders for the AI bots, which until they are set up by an admin with an API key in the Discourse settings are essentially just names of accounts with no ability to do anything, not even respond to a message.
Here is a test message to see if the AI bot responds when not set up with an API key.
However, if the bots are activated then be very careful with who you allow to converse with the bots because if their use violates the policy of the AI company such as OpenAI the account for the API key will be disabled and most likely never be allowed to be reactivated.
As a Discourse admin on another site I too noticed the update and did check out the bots without the API key set. I currently consider them harmless when not activated and unless you have a savvy user, most users will never know about them at present or even how to make use of them.
On this Discourse site I do have access to the bots, so here is a demonstration of the same test.
When a message to a bot is created on this site the follow notification is posted at the bottom of the text input area.
Same test done on the OpenAI forum. (I am a category moderator there)
Here is the official Discourse post
For those without access to see the admin settings here are the default values taken from a current Discourse site
https://<Discourse site URL>/admin/site_settings/category/discourse_ai
Notice that no API key is set. Also never publish an API key, keep them safe at all times. If by accident you do publish one, sites such as OpenAI have a page to revoke a current API key and immediately reissue you a new one.
They donât do anything without the relevant admin settings enabled, but we do appreciate that it would be better if they werenât created unless you actually enabled the AI Bot module. Thereâs a change in the works to make it so, so they should be removed from your site in the near future as that change rolls out on our hosting.
Is that restricted by the trust tier system?
If not this could just be maintained by not posting their names, if that can be customized to different names.
Good question.
I may answer this latter but currently donât have all of the details. I will note that I have been involved in some of the use of the bots for about a month or more now and will note that one should not consider anything as set it stone. Also if you have ideas, suggestions, recommendations etc. please make them but donât expect a reply/implementation or anything else as this is a very very dynamic part of Discourse development.
If you have admin rights to a Discourse site then check out this URL to learn more, that is where I first went after seeing your question.
https://<Discourse site URL>/admin/site_settings/category/discourse_ai
EDIT
I posted screenshots of the current Discourse AI bot default settings into an earlier post so that everyone can see what they are but more importantly have a reference so they can follow along in conversations about the details of the settings.
To answer your question
Thanks, I am an administrator set up API key for this but donât have anthropic/stability keys only one open A.I. key.
The default setting says the only allowed group is trust tier #4.
FYI
I havenât used the bot on this site in a few weeks but did want to point this out as it may not be obvious on first use.
The bot can have different personas which can be selected but the selection option is not so obvious.
If one clicks on Forum Helper it is really a drop down list.
Enjoy!
Also of note is that since the bot is currently accessible via a private message and other users can be invited to a private message, when two real people or more and the bot are in the same PM the bot responds to every post, even it the post was not directed toward them.
Is that limited to trust tier #4 for this support forum?
Iâm not an administrator here just for my own site.
There is this notice posted with that:
AI chatbot messages are monitored regularly by moderators.
Still doesnât seem to work as the default system bot stopped working when I set site to staff-only mode a lot is restricted with that but plan to re-launch next month.
Setup Open A.I. account linked with microsoft account, they give $5 free credit for that but if there is a lot of traffic then they bill for that.
Interesting they donât want tax I.D. number for companies in the U.S. but they do for other countries. 
Obviously I canât answer that based on the settings on this site as I donât have access but will note there was a group setup for early test which is how I was included.
https://meta.discourse.org/g/ai-bot-testers?asc=true
If you look at my examples earlier you will see that there are two different wordings and AFAIK that is this setting
For the site I have admin access we will not be setting up the AI bots. For the sites I have access to running Discourse AI bots I do not have admin access. I also donât plan to set up a dev site and add an AI API key to test this so if you stay active with what you are doing you can carry the baton in a week or so. 
I can not help with that.
Alright good info thanks for your help.
