LLMs enable cheap deanonymization

Researchers built an LLM‑driven pipeline that extracts identity cues from free‑text posts, searches the web for candidate matches using semantic embeddings, and verifies matches — identifying many pseudonymous users (e.g., Hacker News→LinkedIn) at commercial cost ($1–4 per profile) and high precision. The attack works on raw text across arbitrary platforms and outperforms classical deanonymization baselines. — This shows practical anonymity on public forums can be rapidly and cheaply defeated by automated LLM pipelines, forcing policymakers, platforms, and vulnerable users to rethink privacy, whistleblower protection, and moderation rules.