Research
152d agoLLMs Expose Online Identities at Scale
Researchers demonstrate that commercial LLMs can reconstruct user identities from public posts, revealing patterns that compromise privacy worldwide. The method leverages prompt engineering and vast training data, enabling cross‑platform inference.
The Signal
Experts warn that the technique could be weaponized by state actors or malicious firms, citing OpenAI as a prominent example, prompting calls for stronger data‑protection standards and AI‑ethics oversight.