from the mod log:
https://s.faf-pb.xyz/lXxek was the old url but they took it down - I have a copy at https://join.piefed.social/wp-content/uploads/2026/05/evidence-llm-used-for-banning.txt
The output from the LLM was provided as proof that someone needed to be banned.
I didn’t want to do this but my hand was forced by their dissembling, minimization and bullshit. Note how the output includes a nice big logo with a slogan and ascii graphics - this was not a one-off experiment, this is an app they were proud to call their own.
I’ve not followed along all the various threads and, in all honesty, I’ve only read this very post because it popped up as being reported while I was trying to drink my morning coffee. Also, don’t ask me how I ended becoming a mod either, I have no clue.
Imho, and I say this as someone only minding the meaning of words, if those people really see themselves as anarchists I would suggest they urgently:
- Look up the definition of anarchism in a (decent) dictionary and see how well it fares using tools like AI (being a centralized tool that is owned by some powerful/uncontrolled group, the IA itself being a true black box… while they, the ‘anarchist’ users of said AI themselves trust it, and so on). Or maybe, since they’re already using it, they could ask their AI if they’re really meant to go hand in hand? ;)
- Spend some time finding a better suited name to describe their jolly band. Imho, “believers” seems much more suited than “anarchists” to describe any group of people willing to listen to words coming out of some black box outside of their control but that is just an impression, there may be better words out there.
I don’t mind people using AI: it’s a new tool and it is to be expected that some real stupid shit will be done using it. But, like any tool, AI is not neutral. By and in its very design AI has a purpose and if I was to define myself as an anarchist, which I would not if there was still any doubt about that, I would seriously and very urgently question my desire to use AI despite its intrinsic design, despite its very purpose.
Now, if you don’t mind, I will pour myself a fresh and hot new cup of coffee. And maybe later today I will pick up my copy of Proudhon ‘Qu’est-ce que la propriété?’ (‘What is Property’, which was first written in French) which, imvho, any wannabe anarchist should read before deciding it’s ok, or not ok, for them to use AI, and what type of AI. And yes, that is a book worth reading for discussing AI despite having been written in the first half of the 19th century. The answers they might get out of that book may very well… unsettle them and their apparently very unchallenged certainties.
A very stimulating and healthy read, even for non-anarchists like myself.
“I don’t know anything about the situation at hand, the people in question, nor the political theory they espouse, but I won’t let that stop me from stating my opinions as facts with the confidence of a 100 cardinals. Also, read this 200-year-old political treatise book that will definitely apply to bleeding-edge GenAI technology.”
Almost Impressive demonstration of proud(hon) ignorance. Enough with that.
Accidentally unlocked the comment (misclicked when the giant 3-dot menu expanded under my thumb), but I don’t see why it should be locked in the first place.
Re-lock if it’s supposed to be locked, I guess.
Libb locked immediately after insulting db0, I would say it was inappropriate to have locked in the first place.
Super cool stuff. I like how you suggest a book written by a man whose Wikipedia page, under “Controversial positions” describes him as anti-feminist, antisemitic, and proto-fascist. No one having a good faith discussion throws out a reading assignment while making not one single argument or quote from the book. And to then insult people for not reading it (or rather, not having already read it) is really telling. You dont care what people think, so long as they shut up and leave.
I’ve skimmed the book before and I wasnt shocked to be in disagreement with many of his points. He wasn’t a committed anarchist his whole life, and even notions like “property is theft” are circularly incoherent. You can’t steal something that isn’t already recognized as owned by someone else. At the very minimum, a being has complete autonomy of their own body, so again, hardly surprising that a committed misogynist thought bodies were communal property.
Why do we care what a famously obtuse old proto-fascist has to say about contemporary anarchism again? Oh right, because this whole tirade is to deflect from the main topic with ragebait and probably some crypto-fascist propaganda move.
I can reccomend books you won’t read too. In “The Commune Form,” Kristin Ross makes many interesting observations about the formation of communes and their foundational role in all forms of left-wing revolution. From the intro:
My goal in these brief reflections is not to provide a definition of a form that in its contingency, lack of abstraction, and ongoing, unfinished nature could hardly lend itself to such a task. The commune form, as form, does not lend itself to a static definition, unalterable through time; it does not unfold in the same way everywhere around the world. It is not a concept.
Throughout the book (which is also available online) she makes similar cases and discussions: that the commune form (and anarchism itself) are specific instantiations of common practices. Just like a pirate ship is a floating laboratory for democracy, just like every Fediverse instance is an experiment in governance.
I’ll be frank, you really piss me off. I find that smug, patronizing “I sip my coffee while moderating the children” tone of voice to be such an immediate indicator of a person who let’s the petty tyranny of forum moderation go straight to their head. See also, demanding an entire other instance of people read some book in order to satisfy your demands for ideological purity. You sound like the worst kind of condescending reddit mod, a tiny authoritarian loftily demanding the peasants get on your intellectual level. “Read my dense ideological treatise,” says the “teflon man” who doesn’t let ideology stick to them.
Go fuck yourself.
Go fuck yourself.
Did that help you feel any better or any righter? Not that it matters much. Have a nice day.
Dbzer0 operates a user hosted mesh of local AI, there is no big powerful group. It’s a collective of PC tech enthusiasts offering a free alternative to the big corporations.
https://faf-pb.xyz/?ab3c4aae88b64dbc#J8jz6mNac8kRHm5aMXBJQjbUr9pbwMfWrdEaNm1jywhx
You can see the output of the AI here, it’s simply provided a summary of a users comments and offers URL links to view the original comments. There is no issue of AI neutrality because a human is entirely behind this process, and they make the final judgement on whether to act on the report.
I understand it’s fashionable to hate on AI, and I totally get being anti-corporate and against datacentres but when you’re running something on your PC that is completely unconnected from them that argument falls apart.
It’s a collective of PC tech enthusiasts offering a free alternative to the big corporations.
By using a corporate owned AI?
I understand it’s fashionable to hate on AI,
Allow me to quote my own message:
I don’t mind people using AI
Where do you see any hate in that statement, or in the rest of my comment for that matter? Also, not minding something (AI, or not AI-related) should not forbid me, or anyone else, to look at its in a critical manner, right? Or did I miss the episode explaining us they, for some reason, should/could not be criticized?
Whatever, thx for considering old-me as fashionable. That had not happened in many decades.
By using a corporate owned AI?
Qwen is FOSS under Apache License 2.0.
I don’t mind people using AI
My apologies, I must have misread or skimmed over a part here.
There is nothing un-anarchist about using such a piece of technology, especially when it’s being deployed to catch genocide deniers.
My apologies, I must have misread or skimmed over a part here.
Alas, it has become the norm, to skip ‘a part here’.
There is nothing un-anarchist about using such a piece of technology, especially when it’s being deployed to catch genocide deniers.
Don’t waste your time throwing slogans at me, I’m like made out of some kind of ideological teflon and no slogan will stick with me (even the ones I may agree with).
Instead, of telling me what I should think (to be right) about AI and anarchy, have you even considered opening the Proudhon book I mentioned? It’s in the public domain, easy to find. And maybe try to understand why an obviously out of touch old fellow like myself may have hinted at the possibility that, maybe, using AI-powered tools could not sit that well with an anarchist ideology and the way it supposed to stand against some very fundamental notions? Or is it simply that I must be wrong given the fact that I dare not agree with you or with someone you agree with? If that is so, that’s fine by me all I have left to do is to wish you a nice day.
Instead of telling myself and others that we couldn’t possibly be anarchists because our view on the subject are different to yours and suggesting someone read a 270 page book to understand your POV, why don’t you summarise your reasoning?
Is it that AI is able to free up our menial labour?
Is it that is infringes on corporate copyright and steals from them?
Is it that it’s being used in a manner that harms the finances of AI companies who charge for this service?
Instead of telling myself and others that we couldn’t possibly be anarchists because our view on the subject are different to yours and suggesting someone read a 270 page book, why don’t you summarise your reasoning?
Have you noted the many words I put in italics in previous reply? Maybe there was a reason I insisted so heavily on those. And, once again, don’t waste your time (and mine) trying to put words in my mouth that I’ve not said. The only reaction this might end up triggering (you’ve got some leeway) is in me deciding to ignore all your messages from that point on.
As for why I don’t summarize the book for you and expect you to “read a 270 pages book”? Thx a lot for asking.
We live in the age of “summaries”. People, old and young, don’t want to read anymore. They want an instant summary of those books so they feel like they know it without having to actually put any effort reading it. So they feel confident they can get good grades when there is an exam on that book (the only ones to blame here being the educative systems and the teachers, not the kids abusing the system as they’re acting quite smart by doing so). Or feel like they can knowingly talk about that book with their peers.
People who want to use that shortcut are more than welcome to go that road if that’s what they really want, you are more than welcome to go find yourself a summary of the book, even though I think it’s a sad mistake. You can even ask one of the many AI to do summarize it to you (maybe this could give you enough motivation to start reading it?)Just don’t expect me to summarize the book for you for the simple reasons that I think one must be willing to read a book not just act like if they had (it’s sad but it’s ok to not read at all), and then because I think most of the value one can get out of reading lies in the act of reading the actual book, taking the required time to read it and not in the act of getting a ready to memorize summary of it.
Instead, what I’m willing to do, if you ever decide to read it and put in the work, it is to take time to discuss it at length with you, discuss any question or objection you might raise… to the best of my own limited abilities.
Obviously, that is just my opinion, feel free to ignore it.
A user hosted mesh of black boxes that are opaque as to how they made their decisions (yes even the ones that “say” what their thought process is).
Also, don’t ask me how I ended becoming a mod either, I have no clue.
Ah you don’t remember but it was for editing the piefed wiki and the fedihack event. The rest, i completely forgot.
Considering the recent actions, maybe reconsider that.
I saw the post on dbzer0 related to libb’s lock action. I guess, your real question is my position in this mess ? So be it.
I don’t want to take side. It was posted by the Admin of Piefed.social and raised an issue on how we handle the moderation part. This post is also there because someone asked why Rimu doesn’t post in !fediverse@piefed.social.
- I haven’t be able to diggest this enormous raw of informations. I prefer taking my time and read them slowly outside the current debate.
- Both Rimu and Dbzer0 can counter-arguments, expose their facts and users can decide. I’m watching. I will lock the post later so i can take a break.
- As for myself, I would have handled differently.
- I have a mixed opinion on AI. I use it for subtitle or transcript. There are pro and cons. Dbzer0 stated they don’t use it, or run a local one on you computers. I believe them.
- I understand why people would feel their privacy was invaded, i’m one of these. The reason we left Reddit was for spez locking up API and later corpoAI. But our post are public and we check user’s history before taking decision.
- I also understand why mods would use this tool. It is covenient, fast but that’s not how i would handle or imagine the moderation part.
Maybe my opinion would change overtime if i was managing a bigger instance.
As for the discussion between Unruffled and Libb, It’s better to let them alone.
If Libb want to disengage from the current discussion, so let it be. I don’t think it is worth to insist and ask him to discuss when he doesn’t want to.
Maybe they both missed an interesting discussion and misunderstood each other ? Maybe later they could share their ideas, books and define part of anarchy in the 21 century ? Maybe they will completly disagree ? Dunno.
Maybe, they could also imagine a disengage tool where anyone, not only mods, can lock their own threads and tell users “i don’t want to continue this discussion.”
Right now, It is not the right time. It’s not the right place for a talk between libb and unruffled.
So i’m waiting for a better moment and that we cool down.
Not a question, more a comment/feedback.
I consider libb’s reply less disengaging and more enrage and block, which I consider inappropriate for a mod to do.
Regarding the larger bit with rimu and this post in general (not what my comment was about, but I’ll elaborate since you took the time to do so as well), the issue - for me - comes down to veracity.
I think its perfectly appropriate to call out bad actions. I have no problems whatsoever expressing my disappointment with the FAF when it has come up, which has been rare, and expressed those thoughts without any issue whatsoever. One example being a discussion about defederation, where some of the examples shown should/would have been removed anyway due to the rule breaking contents.
Anyway, for me, claims like this should be met with a pinned post that they haven’t been verified, and each user should review and consider accordingly.
For example - is there a tool using an llm reviewing comments at request of a user? Yes, easily proven.
Is it an automod? No. That claim is unverifiable with what has been shared, and not even remotely indicated by what was shared, making the bulk of this post conjecture.
As for the points (I’m going to skip ones I fully understand and agree with to not make this too long, because I have a feeling you are already occupied with keeping an eye out for rule breaking behavior):
have a mixed opinion on AI. I use it for subtitle or transcript. There are pro and cons. Dbzer0 stated they don’t use it, or run a local one on you computers. I believe them.
Same. They definitely run locally, as do I, the ai horde has been a publicly available tool for a long time. I’m very much against corp. ai, but perfectly fine with local use.
But our post are public and we check user’s history before taking decision.
And most importantly, none of that is different here. Everything else in this post is just wild speculations and blatant fear mongering. Which is the issue I have with this post, and why I think a pinned comment is appropriate.
As for the discussion between Unruffled and Libb, It’s better to let them alone.
Again, I was more referring to their interaction with db0.
Flat out inappropriate, no excuses that are valid there IMO.
Edit: Autocorrect typo corrections
Oh, now I understand. Well I would say that’s OK then. Just be aware I won’t do night watch ;)
Same, i will probably lock it during my sleep time than unlock when i’m awake. I did the same with a luanti server.
For now the thread is going well. Thank to them all. :)
And don’t do mod task if you only wanted to do the wiki or something else. ;)
And don’t do mod task if you only wanted to do the wiki or something else. ;)
Oh don’t worry about that ;)
It just happened I considered the OP post oddly interesting to comment on it (even though I still think this is not the best place to discuss that kind of questions, it also opened a few very interesting perspectives worth exploring) and I also wanted to make it clear to the person that 'reported it that it had been dully noted by a mod (ie me wearing my mod hat, in my first reply) and that said mod saw no real issue in the OP post.
After that, the few feedbacks I received had promises in them. I tried to reply as openly and as usefully as I could. Alas, it seems we’re quickly moving towards a more ad hominem type of discussion around the classical “you’re doing it wrong or you must be dishonest if you don’t agree with us/me” type of exchange that I won’t contribute much more to if that doesn’t change. Will see.
Well don’t hesitate to lock with a message that explain why you lock the post if we move toward this kind of exchange.
It can excalate quickly and we have lot things to do. :)
Will do if I have to, but like I told the person I’m more than willing to give them an opportunity to keep this a real discussion, not an ideological pillow fight ;)
If you read it, so far it is still perfectly correct with barely a few hints at… anger, or not even that: a mere desire to trigger some emotional reaction on my part maybe.
FYI, i will lock this post in 3hours at 19h30pm. So we can take a break from it (and myself too).
Thank you.
Edit : Sorry, I’m locking the post and taking a break. I also invite you to do the same. That’s not easy but i really wish that we can find some peace of mind.
I apologize to anyone that were hurt by this whole mess. I couldn’t do better. Sorry.
I feel like the post title should be changed before locking it down because it seems a little misleading based on the comments.
I don’t think that’s a good idea.
Editing the title would rewrite or paint Rimu as someone who tried to hide something or told something else.
I think it is better to let as it is, they will read the title and the comments for what they are. Then like you, they will understand.
Well…First, i want to thank you to you all, the topic wasn’t easy but the discussion is going well.
I hope the thread will remain that way :)
And? It’s in the public modlog, Rimu. Please, can you stop harassing me? How many hater posts are you gonna make this week exactly?
this was not a one-off experiment, this is an app they were proud to call their own.
It’s a python script that calls the lemmy/piefed api to grab comment histories. Calling it an “App” is quite a stretch. Luminous just likes pretty things, bless his cotton socks.
You are right about one thing, we will no doubt keep using the script to pull comment histories on the few occasions when we want a deep dive into someone’s profile.
For the record, we were never planning to use it as an automod or hook it to an LLM. You hallucinated that, ironically. That’s why we never announced it. It was nobody else’s business.
How it was actually intended to be used was as a tool to supplement the existing lemmy mod tooling, so that mods can make better, more informed decisions, in a more efficient way. How is it ever a bad thing to have all the relevant information to hand? And it’s being open sourced soon, so everyone can use it. Unlike some instances [nudge, nudge], we try to take a considered approach to user bans. And no, it’s no linked to an LLM in any way. If you chose to use one, you’d have to paste in the user history manually. But it also very useful just for skimming through, much faster than with Lemmy UI.
Also for the record, we do not use any form of automod with automated banning and never have. All ban decisions are made by a human mod or admin. Neither do we auto-scan user profiles for bad political takes, like Zionism. What we do do, is act on user reports, conduct an investigation and then make a mod decision. Nothing has changed in that respect, except now we have an extra tool to make sure we get the decision correct.
Finally, just because I included an LLM summary of the comment history in the public modlog, it does not logically follow that the ban occurred because of the LLM summary. It simply happened to correspond very well to my own manual review, which I have already explained to you. And I thought the summary did a much more comprehensive justification of the ban than I could have fitted in the modlog, so why not include it?
In conclusion, much ado about absolutely nothing. Again.
Aside from all the drama, is there a reason you are using the gpt-5-mini api from OpenAI (and aparently pay them for that) instead of a self-hosted model? This seems to be also against your instance policy and helps OpenAI to harvest training data.
We added the chatgpt reference because we thought it would be funny, luminous talked about it somewhere already. We are pirates, and self-hosters. This is surely obvious. And we don’t have cash to splash around on corporate AI tokens.
Sadly this is a learning moment that inside-jokes don’t belong in modlogs. Any nuance it would have would be immediately lost and twisted beyond any recognition.
Yeah, this went right over my head so I was confused when I saw mention that you are actually using a FOSS model. Personally, I see nothing wrong with that as a tool to supplement mod work
So what model are you using? Models are biased by their training data and guardrails put in by the trainers, so for transparency sake it is important to mention the model used.
Based on another comment in the thread it’s Qwen which is a FOSS model made by Alibaba
The source is not open so it is not FOSS. It is freeware.
Depending on the exact model, it may be under Apache 2.0 licence, which is FOSS.
Can you point at the source data the model was compiled with? If you can’t, the source isn’t open.
Qwen iirc.
Can you define harassment in a way that includes this post?
Flatworm, looking for clarification. @db0@lemmy.dbzer0.com alleges you didn’t use AI in your ban process. But then I read your comments and they are far more difficult to parse.
Finally, just because I included an LLM summary of the comment history in the public modlog, it does not logically follow that the ban occurred because of the LLM summary. It simply happened to correspond very well to my own manual review
- you claim you had a manual review process, but never show your work. (If you did, where is it?)
- you imply you didn’t use the LLM to ban the user, but you don’t say it outright.
And?
Looking for commonalities by parsing logs is something llm’s are really good at.
What, are you just trying to start another fuck_ai brigade event?
Ok if this is not a big deal why did it fall to me to blow the whistle on it? Why were they so quiet about it?
lmao nothing “fell to you” and no one is “quiet” about anything. you just took it upon yourself to take an already public modlog (one that unruffled knows is public) and follow the link that he posted (pubicly, to a page that anyone who has the link can view) and then pretend you found some big secret. this isn’t whistleblowing, it’s closer to fear mongering or some delusion where you think the world needs saving and you’re the valiant hero.
luminous and unruffled both admitted everything openly and most of the shit in your original post was entirely untrue.
This is a whole lot of absolutely nothing. Thats why.
BREAKING NEWS! Mod team develops and uses mod tool!
„Anarchist” mod team delegating work to a non-deterministic tool operated by big tech is the news here. Kind of confirms my understanding of who calls themselves anarchists these days - people who value freedom to consume and freedom to be exploited above actual freedom.
There is no delegation to the tool.
It provides an evaluation in text format, and references. Thats it.
Edit: To be clear, rimu is just making that part up. It has no ability to auto-ban and its not an automod.
Its a python script to check a user’s comment history and evaluate using a local model, and spits out what it finds. A person has to read it and make a decision on this information.
Everything else is rimu’s hallucination.
If there was a full review of source text that was summarised by an LLM then it wasn’t needed. It’s a pointless tool that makes guesswork look very confident. This guesswork was done by a black box ran by OpenAI.
If there was a full review of source text that was summarised by an LLM then it wasn’t needed
You go through every single users comments?
Do you realize how long that would take?
I disagree, its a huge time saver.
This guesswork was done by a black box ran by OpenAI.
Local qwen, summarized with receipts.
https://faf-pb.xyz/?ab3c4aae88b64dbc#J8jz6mNac8kRHm5aMXBJQjbUr9pbwMfWrdEaNm1jywhx
Is the evidence in question that Unruffled acted on when deciding to ban the user for breaking their instance rule against Zionism.
Your previous post made it sound like this was an automatic AI going around banning users for wrongthink, when the reality is its a search tool for mods and admins to quickly look through a users potentially years long comment history.
It is chock full of examples of such behaviour, along with hyperlinks to the offending comments. What exactly is the issue here, that an AI tool searched through a log and pointed out examples of rule breaking behaviour and got a person who downplays genocide banned?
It is chock full of examples of such behaviour, along with hyperlinks to the offending comments.
The ban log makes it unclear whether the comments are all offending, or even supposed to be. Was the user banned for this?
Date: 2026-04-17T13:51:01.747786Z Comment ID: https://sh.itjust.works/comment/24890311 Post ID: 591643 Community ID: 351 When the Molly hits just right.
You likely would have to use some moderation tools on a public forum. How else do you even attempt to stem bot accounts?
I think it’s notable that the big receipt reveal is an example of the LLM mod tool absolutely nailing the assessment.
Honestly - this type of reporting for ban rationals should be the gold standard of moderation, AI or no. If an AI tool can help raise the standard for transparent moderation, I am 100% in support of it.
Civilian de‑protection rhetoric Notable examples:
- Claims that “minors are not innocent by default”
- Statements that “they’re killing children” is mere outrage propaganda
- Questions implying all Palestinian deaths may involve combatants
Yeah I’d like to see direct quotes but even if this is heavily paraphrased there’s no question.
Date: 2026-03-15T10:53:32.311868Z Comment ID: https://sh.itjust.works/comment/24299783 Post ID: 519441 Community ID: 45
Just because they are minors, doesn’t make them innocent by default.
The whole article is devoid of details on how these drones are used.
Control-F is your friend
I didn’t realise how long that page is, motherload of receipts
Full comments with direct links are there, thats the summary up top, just scroll down.
Edit: https://faf-pb.xyz/?ab3c4aae88b64dbc#J8jz6mNac8kRHm5aMXBJQjbUr9pbwMfWrdEaNm1jywhx if you want to see for yourself.
Don’t be dishonest. You know there are no direct quotes there.
The quoted comments are directly quoted with links.
You are lying. The AI Sloppary says
Claims that “minors are not innocent by default”
Now show us exactly where this exact quoted text in the comment dump. With links, the things you swear you have!
========================================
Date: 2026-03-15T10:53:32.311868Z Comment ID: https://sh.itjust.works/comment/24299783 Post ID: 519441 Community ID: 45
Just because they are minors, doesn’t make them innocent by default.
The whole article is devoid of details on how these drones are used.
========================================
Whats the prob Bob?
Look at that, the words don’t match what’s in the quote marks, and it has a slightly different meaning. My whole point.
I’m almost impressed that you can lie this hard while acting correct.
Unruffled also doesn’t tag AI posts, violating db0’s own AI policy. But db0 doesn’t seem to care. The place is getting more and more obnoxious by the day it seems.
Wtf is this coordinated assault on dbzer0? Even your own links are just pointing to Flatworm very reasonably responding with their exact moderation process. I know it hurts to be wrong baby, but when someone writes you a short essay addressing all the points and then some, that’s not “minimizing” your complaints, is disassembling and disproving them.
It’s weird because this thread sits right above the one from a week ago where its detailed how people were trying to fake neo-nazi content against the db0 instance.
Like, context is important here. The piefed dev knows exactly whats going on and who these individuals are, especially the user in question that got banned.
It’s an ugly look for sure.
Dbzer0 is full tilt groupthink and exclusionism. They consistently follow their leader and error-prone tech to aggro any form of defection— it’s hard to parse it from a cult. Rimu stays based for calling out them assuming guilty before proven innocent. I only wish the Piefed full instance block feature still worked on the Voyager app.
Rimu stays based for calling out them assuming guilty before proven innocent
Can you clarify please?
The tool summarizes, includes references to specific comments, and a person has to review and action.
What “assumed guilt” are you referring to here?
As long as the moderation follows their rules, and it is always as transparent as shown in this example, I don’t see an issue with this.
My only concern is that LLMs are very good at recognising biases in questions and are more likely to confirm them than push back. So the LLM might pay too much attention to small/possible infringements. But this depends heavily on the model, the prompt, and the reader.
The output from the LLM was provided as proof that someone needed to be banned.
Alright, but isn’t that a whole lot more than most people getting banned or deleted would typically get? Slop or not, and AI opinions aside?
