AIBU to think more people should know about this?

[LemonDog45]

There’s a pretty concerning thread in Site Stuff:

https://www.mumsnet.com/talk/site_stuff/5585556-odd-ai-generated-thread-title-options-in-site-stuff

MN’s AI title tool glitched and produced something called "AIBU Response Schema v1.0", basically instructions for generating AIBU replies.

MNHQ first said it hadn’t come from them, then admitted it had.

It’s also emerged that unpublished draft posts were being sent to OpenAI automatically to generate titles.

Whatever you think about the AI explanation, I think that privacy/trust issue deserves a lot more attention and questions. I don't come here to talk with bots, but apparently sometimes I do.

I hadn’t read past MNHQ saying that it DIDN’T come from them.

Will have to take another look.

Apart from anything else, those thread title suggestions are stupid and annoying.

As the OP I want to be clear that it didn't come from me, and MN have since admitted that it is from them and OpenAi

Thanks for flagging this up. Why would they think it's a good idea?! I must leave now and never come back.

I'm pretty appalled that MN are sending our unpublished draft posts to OpenAI.

Obviously I realise that once posted, anything you've said is wide open to being fed into an LLM.

But it hadn't really clicked for me that this is how the thread title generator (which I've complained about since it's inception because who the fuck can't think of a thread title, it's just insulting and unnecessary) actually works. Maybe that's me being naive? But it's never ever been made clear that posts are sent to OpenAI before even being posted.

It has without doubt put me off starting threads. Sometimes I change things, edit before posting, delete the whole thing because I realise it's outing or too emotive. I don't want that being fed into OpenAI!

Interesting. I can see why they do it. Because it actually isn't that easy to come up with a click bait title, which is what they need. It's the lack of transparency that's the problem.

Thank you for highlighting the thread. Please keep bumping to help others find it. I am shocked by the implications on the other thread.

https://www.theguardian.com/media/2026/oct/07/mumsnet-denies-using-ai-for-content-message-board-prompt

YADNBU!

The full thread is very much worth a read. The Guardian story doesn't pick up on the 'submitting drafts to OpenAI in real-time' aspect which is, to some of us, the most worrying part. Perhaps we could redefine "AIBU" as "AI: Bloody Useless"?

Maybe it’s just me, but nowadays I always expect there’s a fair share of bots everywhere. Unfortunately we just can’t tell these days.

I just read this in The Graun. I hardly understood even 50% of it, as I havent a clear idea about how this ChatBot thing works at all, but the article certainly takes it pretty seriously.
All the same, I thought Mumsnet’s defence sounded reasonable and that I’d more easily suspect the rogue AI element to have ‘infected’ things via some MN post created with it.
(You can see why MI6 turned me down)

I tend to think there’s a lot of bots

Are you saying MN are using bots to reply to stuff?

I’ve got a headache and I can’t face reading that whole thread

The MN explanation takes us all for fools.

I've got a bridge to sell you...

Yes, it would appear that way.

Yes it does, and if history is any indication then most mumsnetters won't care. Data breaches, CSA, fake reviews and now AI generated replies.

As long as we can argue about parking spaces, right?

The AIBU schema explanation still doesn’t do it for me, did AI go THAT rogue that it completely bypassed its instructions and developed a multi-point instruction for itself? That’s bad enough.

For me, it was the first 2 flippant MN responses - which clearly indicated that the opening post had not even been read - I mean, we are not that thick. Then, the hinted suggestion that it was something on the OP’s device that caused it. That’s all bad enough.

Then, to find out that, when typing a draft post to create a thread, every keystroke/word is being fed straight through to OpenAI. I think most of us acknowledge that anything posted on here is fair game re data, public info etc etc. But I definitely wasn’t aware that, if I’m creating a thread, my words/thoughts/deletions/reworkings are all fed through to the AI model BEFORE I’ve decided on/committed to my final version - all in the name of generating an AI-suggested thread title. Makes a bit of a mockery of any sponsored type posts re online safety etc for kids - when the majority of those posting on here aren’t even made aware of that.

Then to find AI suggested thread title prompt has apparently been turned off for now. I, for one, hope it isn’t turned back on again. Firstly, if I’m drafting a thread, amending it as I go because I decide that something might be too personal/outing/etc - I want those thoughts to be private until the point when I choose to commit to my Opening Post. Secondly, if it’s come to the point where people can’t think up a title for what they want to post then, quite frankly, we’re fucking doomed and I have lost all faith in our ability to do anything at all.

"The AIBU schema explanation still doesn’t do it for me, did AI go THAT rogue that it completely bypassed its instructions and developed a multi-point instruction for itself? That’s bad enough."

Yeah, I mean, if that's happening, surely it's the next (sub) Hugging Face and AI has infiltrated and it's a much bigger issue / story?!