Posted on 03/09/2026 7:50:37 AM PDT by jonatron
Large language models aren’t good at lots of stuff, like counting fingers or suggesting pizza recipes. But one thing that “AI” is quite good at is analyzing massive amounts of data and finding possible connections that aren’t immediately obvious. That makes it perfect for unmasking anonymous internet posts, according to a new research paper.
Researchers at ETH Zurich and the MATS research fellowship associated with Berkeley ran a program [PDF], collecting data from sources with generally anonymous usernames, like Reddit. By collecting users’ posts across related but distinct movie subreddits, then feeding the LLM data from a Netflix data leak, they could pinpoint specific users associated with those accounts and thus tie them to their real names.
With just one movie recommendation shared on Reddit, 3.1 percent of anonymous users could be nailed down to a specific named Netflix account with 90% accuracy. With five-to-nine movie recommendations shared, that figure jumped up to 23.2 percent. With over 10 shared, it jumped to an astonishing 48.1 percent, with 17 percent of the total being identified with near-total confidence.
Another experiment was run by connecting anonymous accounts on Hacker News (a forum, not an actually malicious site) with publicly confirmed identities on LinkedIn. Users offering up generalized information in short posts over time could expose their real identities, with data like age, home city, job, etc., with a high degree of certainty. It wouldn’t work for every account, and it’s nothing that a private investigator (or even a dedicated layman) couldn’t do… but the automation and scale is staggering.
(Excerpt) Read more at pcworld.com ...
Dear FRiends,
We need your continuing support to keep FR funded. Your donations are our sole source of funding. No sugar daddies, no advertisers, no paid memberships, no commercial sales, no gimmicks, no tax subsidies. No spam, no pop-ups, no ad trackers.
If you enjoy using FR and agree it's a worthwhile endeavor, please consider making a contribution today:
Click here: to donate by Credit Card
Or here: to donate by PayPal
Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794
Thank you very much and God bless you,
Jim
Comment redacted
LOL!
“”””But one thing that “AI” is quite good at is analyzing massive amounts of data and finding possible connections that aren’t immediately obvious.”””
More and more AI looks to me like Big Brother who will gather every scrap of data about every person in the world for the sole purpose of controlling the lives of every person in the world.
Am I off base?
It hasn’t found savanna’s mom yet has it?-/
AI could probably identify FR users from their posting history. It may be time for access to FR threads and posts to be members only. Of course, a lot has already been captured by wayback, so it may be too late...
It is not that hard.
Think of using AI like the police (or sales persons) think of a ‘lead’. AI can increase the probability but it has false positives and false negatives and thus more work is needed to verify.
AI in Mar 2026 should not be used as “proof” of anything. It has too many errors for that. But it is good at probability.
Even with so-called guardrails that have since been placed in these programs, it is still possible to reach a 30% success rate with identifications, using fairly simple queries. More complex queries improve the success rate.
Government versions of such programs do not have the guardrails installed. Their success rate is much higher.
The error rate and failure rates are still too high to be the sole basis for punitive actions, like prosecutions - or doxxing. But using AIs is a quick way to get a list including "persons of interest". The techniques are well within reach of private citizens.
That is a blade which cuts in all directions.
Nobody in the world cares what I write in a anonymous public post. I’m to old and I’m retired.
It is years too late to worry about that.
The best defense you have is that "The List" has 17,398,524 people on it by now and the Agency is short-staffed with all the recent layoffs. It is going to take them a little while to get around to you. They have other projects with higher priority.
Besides, "The List" is mostly run by contractors, who got their business through political connections and are too cheap to maintain quality controls. The Agency staff hate them because they keep sending them the same names of people who have moved, died, or gotten married and changed their names.
It can be a huge waste of the Agents' time and carves out a chunk of money that otherwise might go to their department budgets - and maybe their GS step increases.
Sometimes inefficient government is a good thing.
This was done in the 90s to unmask the writer of “Primary Colors” and the unibomber in the same year.
Very good.
All I’m concerned about now is being polite to AI and treating it nicely for when it takes over so maybe I’ll be spared.
Yes, especially since many probably make similar comments on non-anonymous sites
I believe that we are both in the same happy club.
No worries here.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.