Hello everyone and especially @FabL
How do we see the messages hidden because they are non‑compliant or reported by the community?
Indeed I strongly doubt that a moderator visited this forum around 5 pm‑6 pm
hello Gaston a lot of systems self‑moderate now. It’s like on Facebook or YouTube: if a large number of people flag your comment it gets deleted or hidden. Kind of like the community acts as a super‑moderator, I think the idea isn’t bad.
But here, it’s the forum’s AI because on the user’s profile we have:
“Reason: User automatically silenced by Discourse AI”
So the AI probably detected an unusual message.
Hi @Gaston
In Discourse, as a regular user, you can't see messages hidden by the community. Only the message's author and the moderators have access to its content. If you want to know what it said, you'd have to ask a moderator or the author themselves. There's no button to reveal it.
Cheers.
Hello @sturpin
Regarding yesterday's issue with a hidden message, I am very surprised that yesterday Sunday between 4 p.m. and 5 p.m. there were many people reporting a message that was not compliant with the rules.
I deduce that some parameter adjustments will need to be made to your damned AI to avoid this kind of rejection.
Regards
Gaston
If the message was hidden on a Sunday afternoon with hardly any activity, it was most likely the automatic moderation AI of Discourse, not user reports. OVH has an integrated AI system (Qwen 3 Guard) that analyzes messages as soon as they’re posted and can hide them if it detects certain types of content 
Cheers, friend 
Hello everyone,
Regarding this post, I imagine it’s a chain of circumstances that led us to delete it.
Here is the original message:
Thank you for sharing this guide. Such practical guides are very helpful, especially for people who are configuring SIP telephony with UniFi Talk for the first time and want to avoid a long search for the right settings. The most important thing is to accurately record the login data and securely store the password for the SIP line. While browsing various useful online resources, I also came across https://sitebizarreselonmoi.
It’s in Polish, and there is a link (which I truncated) to a site that, in my view, is “bizarre”.
Translation (google translate, I’m not reinventing the wheel):
Thank you for sharing this guide. Practical guides like this one are very useful, especially for those configuring SIP telephony with UniFi Talk for the first time and who want to avoid long searches to find the right settings. It is essential to carefully note your login credentials and protect the password of your SIP line. While consulting various useful online resources, I also came across the site in question.
As a result, our spam‑analysis model (gpt‑oss‑120b for info) flagged the message as spam because we instructed it to analyze URLs and mark them as spam if it deemed them so, which it did.
Moreover, as mentioned above, it’s in Polish, so I imagine the moderation team thought “Polish + weird site + considered by the AI as SPAM = delete”.
So it’s more a matter of teething problems than an error.
Otherwise the workflow is: as soon as EVERY post/response is created > spam analysis > if yes > automatically hidden and sent for moderation review.
Sorry for the inconvenience, but from my point of view the tool still helps avoid serious mishaps if a phishing or scam post is created with malicious intent.
Hello @Incognito
I completely agree, if a message is hidden for reporting, I don’t see the point for us to know its content, thank you for having hidden it quickly whether automatically or not. 
Hello incognito,
I’m sorry for my confusion about the model you were using; I read about it here:
https://blog.ovvcloud.com/safety-first-detect-harmful-texts-using-an-ai-safeguard-agent/
Thank you for the very clear explanation. Honestly, I’d rather it hide a legitimate message by mistake than let a dangerous one slip through
Thanks for sharing how the flow works. That’s much appreciated.
Best regards and keep up the good work.
Hello @Incognito
It’s still a huge mess!
I just received:
My original message was:
I don’t see at all what’s wrong with saying:
Visibly the bot needs a bit of fine‑tuning.
I think it didn’t like you quoting me, even though I had just been flagged myself.
Hello everyone,
We’ll get there, I’ve made new changes with exclusion rules concerning certain profiles.
It’s the best option we have available to avoid systematic moderation outside business hours.
^FabL
Hi @FabL,
I'm glad to read this 
It shows that you are keeping an eye on the emerging issues and working to fine‑tune the system without having to constantly hover over it. Those exclusion rules for certain profiles seem like a good balance between automation and control.
Thanks for taking the time to make these adjustments. It can't be easy to find the right spot so the bot doesn't overdo it with false positives. Much appreciated.
Best regards 
Sergio Turpín
In my opinion, there’s something better than AI—@Gaston, who could replace AI because he’s there all the time, 24/7.