For a few dollars per target, LLM agents can now link pseudonymous accounts to real identities using only public text.
In “Large-scale online deanonymization with LLMs” (arXiv:2602.16800), researchers from #Anthropic and ETH Zurich show:
- 67-68% recall at ~90% precision linking Hacker News accounts to LinkedIn profiles
- 52% recall for pseudonymous Reddit academics
- 38% recall at 99% precision when matching one Reddit user across time
Comparable classical methods achieved near-zero recall.
The pipeline is simple:
- Extract identity-revealing traits from posts
- Search candidate profiles
- Use LLM reasoning to verify matches
No breakthrough cryptography. No private data. Just scalable inference over public text.
“Practical obscurity” was always fragile. Now it is economically trivial to defeat.





