The LLM Deanonymization Problem: When AI Knows Too Much

Researchers have demonstrated that large language models can strip anonymity from datasets once thought to be safe. Here's what it means for privacy at scale.