Identifiable Information

Personally Identifiable Information (PII) leakage from large language models (LLMs) is a significant concern, prompting research focused on improving its detection, removal, and obfuscation. Current efforts utilize techniques like fine-tuned LLMs, transformer-based architectures, and graph convolutional networks to identify and either remove or transform PII while preserving data utility. These advancements aim to mitigate privacy risks associated with LLMs and improve the responsible development and deployment of these powerful technologies, impacting both data security and the ethical use of AI.