AI Models Can Now Unmask Anonymous Social Media Users With Up to 90% Precision, Researchers Warn

Researchers warn that AI large language models can now deanonymize pseudonymous social media users with up to 90% precision by browsing the web and analyzing free text, raising urgent concerns about government surveillance, corporate profiling, and targeted scams.

• New research reveals that large language models can deanonymize pseudonymous social media users at scale, achieving recall rates as high as 68% and precision rates up to 90%, far surpassing traditional methods.
• Unlike classical deanonymization techniques that require structured datasets, LLMs can browse the web, analyze free text, and use simulated reasoning to match anonymous posts to real identities across multiple platforms.
• Researchers warn that these capabilities could enable governments to unmask critics, corporations to build hyper-targeted advertising profiles, and attackers to launch personalized scams, urging platforms and LLM providers to implement safeguards like rate limits and anti-scraping measures.