Garnet is a generative AI enterprise firewall designed to secure interactions with large language models (LLMs) such as ChatGPT. It enables organizations, including highly regulated sectors like government and eHealth, to utilize AI technologies while maintaining strict data privacy standards. The solution leverages a Qdrant vector database and operates entirely within a confidential enclaive cloud using confidential computing measures.
To protect sensitive information, Garnet processes documents before they reach the LLM. It vectorizes the input, applies pre-filtering, generates prompts, and pseudonymizes the data to ensure confidentiality and regulatory compliance. This architecture prevents external models from accessing raw sensitive data during processing.
