Researchllmde anonymizationprivacysocial media

LLMs De-Anonymize Users From Anonymous Posts

schneier.com

|March 2, 2026

9.1

Relevance Score

Researchers report on March 2, 2026 that large language model (LLM) agents can deanonymize users from a handful of anonymous online posts across Hacker News, Reddit, LinkedIn, and anonymized interview transcripts. Their method infers location, occupation, and interests, then scales web searches to link profiles among tens of thousands of candidates. The finding demonstrates practical privacy risks from unstructured text and automated reasoning.

LLMs De-Anonymize Users From Anonymous Posts

More AI & Data Science News

Single-Session Interventions Produce Small Monthlong Depression Reductions

AI Agents Accelerate Scientific Discovery, Pose Risks

Apple Shifts Siri Hosting To Google Data Centers

xAI Installs Noisy Turbines Near Community

Scoring Rationale

Sources