Anyone who didn't think this was coming was a fool. AI now one-step doxes anyone based on your writing style.
(media.kotakuinaction2.win)
Comments (60)
sorted by:
The method presented in that screenshot isn't "writing style", it's literally cross referencing if you were stupid enough to mention your place and duration of work, city of residence and niche hobbies.
Only retards needed AI to help dox you if you were that bad at OpSec.
For the record, my niche hobby is methods of sealing wooden doors.
That's right jolly interestan, mate. Me 'obby is 'andcraftin' yeeyee-ass shit-tier birdnest shoes. Like'n da fishtank shoes, but fer birds. Wherefore this ideation? Mine bud Bubba livin next town over in Johannesburg said he learned it while he was workin' downin Sydney.
I got in to the business of selling free-standing chimneys.
Now we just have to invent crematorium ovens 18,000% more efficient than the current top of the line model.
My friends Is plumbing for giant swimming pools.
You're not looking at one account and deciding whether it matches a known profile, you're looking at hundreds of millions of accounts and trying to figure out which ones (if any) match that profile. It's an automation problem.
I'd assume that was buried deeper in the actual article (searching for link now). In the meantime, they've already got your info and writing style having scraped your emails for content. https://techcrunch.com/2026/02/18/microsoft-says-office-bug-exposed-customers-confidential-emails-to-copilot-ai/
Joke's on them. I don't send emails that have anything other than copy-pasted text from others.
Joke's on them. I don't send emails that have anything other than copy-pasted text from others.
Joke's on them. I don't send emails that have anything other than copy-pasted text from others.
Please, all my communications are hand mailed letters made from newspaper cutouts. Git gud skrub
Hear hear, if you’re stupid enough to publicly list where you work, what you do and where you live, even in tiny slices across different posts, that’s your own fault.
And yes, I have some personal accounts where I do reveal SOME loose personal info (yknow when you’re posting in a city group about what local restaurant you like). But I don’t cross pollinate those accounts with accounts like this.
Lmao, I remember 15 years ago reading articles about how it only takes something like 15 or 25 pieces of information to identify someone. Stuff like, what music they like, what games they play, etc. People are that unique that in a world with billions of internet citizens, just a relative few pieces of information can identify you.
https://archive.nytimes.com/bits.blogs.nytimes.com/2015/01/29/with-a-few-bits-of-data-researchers-identify-anonymous-people/
This article claims that only 4 pieces of information are enough to identify 90% of people.
This is why I try to say things like "my spouse" and not my wife, etc. This site uses taboola ads (scored.co), so posts are being analyzed/catalogued that way.
But yeah, cant be to many retards like me, who believe aliens created humans, and was dumb enough to get poisoned with refrigerant/refrigerant oils. I think the doctor ive been seeing might end up moving away ( i just have a feeling ), and at that point Im not sure what I will do.
I really dont wanna plaster my face on a article somewhere, but if signing up for maid and going to a reporter and telling my story is what it would take to get some fucking help, then I should.
But to OP's original point, yeah we should all be worried about the surveillance state being built around ai.
Israel's greatest weapon yet.
Israel's greatest weapon is still the jewnited states of jewmerica.
Here's the link OP couldn't find: https://xcancel.com/alex_prompter/status/2026951410676621581#m. And the actual research: https://arxiv.org/pdf/2602.16800.
It wasn't about your writing style it was about information you give in your comments that can be pieced together. But I'm guessing AI is already pretty good at guessing whether two accounts are run by the same person based on writing style alone. Write differently from how you would naturally if you don't want to be ID'd.
That's why I make sure to fabricate details on the regular.
Fabricate? No, I totally beat a leopard to death with nothing but a rock when it tried to eat me. I definitely
lovelive somewhere leopards exist. Promise.Was there recently a furry convention in town?
Not that anyone can prove. Strange occurrence of flying rocks whenever someone tries to bring it up.
This is probably one of those fabricated details. If not you are exposing yourself and should delete this comment.
There's no way to tell which details are correct. The important thing is to encourage as many people as possible to poison their data to ruin the tracking models.
Im just going to start blogposting scenes from movies and TV shows which will obviously throw up matches but the wrong kinds they are looking for.
Sheeeeeeeeeeit I already be writin like dey tells me in school n all dat shit nigga
The strategy is to pass all content through an LLM to standardize it into a professional AI tone, reducing the chance that another model can recognize your stylistic fingerprint.
There’s a fascinating cat-and-mouse dynamic here. Authorship detection models look for statistical quirks—sentence length distributions, punctuation habits, word preferences, syntactic rhythms. You’re proposing to smooth those signals into something closer to the mean. In theory, that reduces signal. In practice, models sometimes learn to detect the smoothing itself. The universe has a sense of humor like that.
If you want, I can tune the rewrite for a specific audience—casual, academic, corporate, slightly conspiratorial, etc.—and we can explore how tone-shifting affects stylistic fingerprints.
And be sure to poison your data with occasional red herrings.
This one time, at band camp...
Very AI coded last line (was that the joke? I may be autistic)
But there is an interesting arms race between Ai-detecting AI, AI to fool AI detectors, AI detectors to detect anti-AI detecting AI...
the entire thing was AI haha. But yeah, that's the joke.
I had an idea of a skit where two office workers use AI to respond to each other's work emails and at the office party both of their AI when the user asked for more info about the other guy tells them: they're workplace best friends and both went along with it thinking the other isn't using AI and didn't want to get caught.
Governments have been able to do this for decades. Everything has a back door. Everything you do is known, always.
Yeah, we found that out years ago with the PRISM program, even before that with Hepting Vs AT&T
I'm pretty sure ZOG already knows who we are
We are simply too numerous for the regime to utilize their resources to track us down and kill
Who says? They’ve been doing it for decades now.
... for now.
Not surprising that they could connect accounts like this but presumably you also need at least one account that isn’t anonymous for them to connect it to you personally.
Also I love the “anonymity era is over”. Hate to break it to the guy but it’s been over for a while now for the large majority of people. To be fair most people don’t care at all about their online privacy
Anonymity only applies to people in your real life circle. To the authorities and big tech, it's non existent.
I can’t go changing my writing style and all that. Dude, they’re building the beast system and I’m not joining that bullshit even if it means I can’t have a job or a place to live anymore, I’ll be homeless. Fuck it. People who think they’re going to avoid it by being anonymous are out of their minds. Nothing is anonymous as it is right now, but even if you could somehow get around the back doors that NSA and unit 8200 have in every piece of smart tech, eventually you won’t have computers or smart phones, you’ll have devices that won’t work at all unless your unique ID is connected to their AI system first.
It's true, it's only a matter of time. This AI/LLM technology seems to be a key piece to the puzzle.
https://communities.win/c/MGTOW/p/1994faNq65/request-a-rumble-channel-and-a-w/c
c/privacy's pinned post has crazy stuff on how AI and human researchers can track you based on your writing style (stylometry). Your word choice, grammatical and syntactical errors can be compared with material known to be written by you and used to doxx you.
AI can do this on its own. You can be identified even if you are not a big shot.
We are not cybercriminals but feminist researchers feel it's worthwhile doxxing and tracking us since we are a 'threat'. MGTOW is all about peaceful disengagement from the rigged system and not playing along with the scam. There is no violence.
If you live abroad and are a somewhat non anonymous MGTOW, they can identify your (anonymous) Rumble accounts and Scored accounts by analyzing your writing style and active hours pattern. Then they can use a private detective agency or local feminist group to send some male leftists pretending to share your views at some place you frequent in real life. There will be hidden cameras recording you as you become friendly with that person and open up. The goal is to 'understand the evolution of your disgusting misogyny'. When the documentary goes on YouTube, you will face problems finding and keeping a job.
Leftists have done this everywhere against right wing groups, and there are documentaries about the same. They will soon target MGTOWs, if they are not already.
If you are based in the US and are a native speaker, please start a Rumble channel where you explain stuff with PowerPoint or LibreOffice Impress slides. Use OBS studios (free) to record your screen and there are many easy-to-use, free video editing software you can use.
You can explain how rubber and eraser are different, the types of punctuation unique to a location ( ( ' ) ' for smiley face in Russia versus the French adding a space before the exclamation mark) and even analyze Reddit comments (with username redacted) and tell us what they give away.
You can also provide internet privacy tips and tutorials to use the web anonymously.
You don't have to show your face and can even use free AI tools to provide realistic voice-overs in natural-sounding English. You will be able to get financial gain if you do Rumble, and if its monetization is enabled in your country.
You can have a broadcast stream running for days where it's just old videos running in randomized loop, people will find you from Rumble's livestream section and even donate Rumble Rants. Existing MGTOW channels can let their viewers know that all future pre-recorded videos will only be uploaded at a certain time like 10 pm, and as a livestream debut. You will get more Rants and more bang for your buck.
Joke's on them, my grammar and syntax are so bad most of the time I'm winging it and will flip between options for too readily.
Playing by their rules won't help. Charlie Kirk never advocated violence against them and they murdered him in cold blood (and celebrated, and called for more killings)
Maybe di same kind consequence wey dem give other people suppose reach dem too.
Di same thing wey dem do, e suppose return back to dem for where their organisation dey, and make e clear who dey work for dem and who dey fund dem.
Feminists lol
Say it aint so, and for the last 30 years Ive thought it was the elite rich who control everything like shadow puppeteers.
You know what's really good at imitating writing styles? LLMs.
It's the same concept as the face scan ID nonsense. The technology to detect is a mirror of the technology that generates and transforms. In the true information age that we are entering, any information can be created or recreated.
I don't even bother being anonymous
I love my country of Botswanan and cheerfully rejoice for thous time of ages has finally bestowed upon humankind cultural enrichment of all peoples into the west. Free Israel from its oppressors and all of Islam. Praise be to hashem. Peace and security and prosperity is upon the world finally in sight
Dissin be AI alarmism. It scrapin stoopids wit stoopy OpSec. Da OP ebin say the example dey use for AI to be findin' it. If I go:
"Hey, I live in Jonestown, go to Jacobin's for lunch every week, and work under Jorge in the Jipping Department of Jack's Electronics"
Den yah, AI can guess, wit only 67% accuracy, who I be. Ooh, scaryterry.
That's the way I was thinking. I don't even want my alts on .win to be linked to one another, let alone myself
Am i the only user on this site without alts :|.
That's a silly thing for one of my alts to say!
Have your different identities talk about different things.
Identity A: Man, just got back to New York to meet family, shit's freezing. Where's global warming when you need it?
Identity B: London right now is unlivable, too bad the bosses at Barclays have a nice offshore haven but refuses to move me to another department...
Maybe I am too low brow but I don't think I could intentionally alter my post enough post by post to beat this thing.
If you ran all of your posts through an offline LLM and told it to write it in distinct voices each time, could you beat AI with AI?
It'n be mighty possible, but iff'n you don't mind me parlayizing a theory, I'm be thinkin' that game recognizes game, and AI sees AI. So it seemings real plausible-like that Robo-Man will note y'all's down as usin' an AI, as part of the dataset.
Better off just purposefully introducing junk data into the scraper. That's what my dad Peter always taught me before he died back in 2021 from the coronavirus. Obviously, don't just do it in a facetious query, but sprinkle pointless falsehoods everywhere that don't change your post or point, only harm the data scrapes.
Bob Kelso from Scrubs: "Stop speaking like a farmer."
Forsooth, mine verbiage has dalliances with all manner of niche lexological dialects.
Well good thing I never made any social media accounts and put samples of my writing tied to who I actually am out there.