Show HN: Hexlock – Replace PII in text with fake data that has the same format (github.com)

Hexlock has introduced a new tool designed to protect personally identifiable information (PII) in large language model (LLM) workflows. This innovative solution anonymizes sensitive data while maintaining its format, enabling LLMs to process the information without ever directly accessing the original PII. After the model generates a response, Hexlock allows users to "rehydrate" the results with the original data, ensuring that sensitive information remains secure throughout the interaction. The significance of Hexlock in the AI/ML community lies in its potential to enhance data privacy and compliance with regulations like GDPR. By integrating Hexlock into LLM pipelines, organizations can mitigate the risks associated with handling sensitive information, all while preserving the necessary formats that LLMs depend on for accurate processing. Hexlock offers both ephemeral and persistent modes for anonymization, allowing developers to choose between one-time sessions or secure long-term storage of anonymized data, thus expanding its utility across various applications while maintaining robust security protocols.