AI Can Now Doxx Your Anonymous Accounts? Here's What’s Going On
A recent study on “large-scale online deanonymization with LLMs” shows that AI can often identify people behind pseudonymous accounts by combining public posts, web search, and reasoning. Researchers used a four-step pipeline: extract clues from posts, search likely matches, reason over the best candidates, and calibrate confidence. In tests, it identified many users correctly, including 226 of 338 Hacker News accounts linked to LinkedIn and at least 9 scientists from interview transcripts. The main takeaway is that AI can make doxxing faster and cheaper, sometimes for only a few dollars, without hacks or data breaches. The results are not proof that every anonymous account can be cracked: the study used controlled setups, and performance dropped with larger candidate pools. Still, it suggests practical online anonymity is weaker than many people assume, especially for anyone whose old posts reveal recurring personal details.
