Cohesity secures US patent for AI system enabling secure GenAI on secondary data

Cohesity has received U.S. Patent No. 12,619,501 from the US Patent and Trademark Office for technology used in its AI platform, Cohesity Gaia. The patent covers a method designed to enable the use of generative AI with secondary enterprise data while maintaining security and governance controls.

The patented invention, titled “Data Retrieval Using Embeddings for Data in Backup Systems,” introduces a retrieval-augmented generation (RAG) semantic layer. According to Cohesity, this layer allows generative AI applications to access and use secondary data without requiring additional data silos or exposing sensitive information. The approach is designed to keep data in its existing location rather than creating separate copies for AI processing.

The development of the patented technology involved Gregory Statton, Sanjay Poonen, Mohit Aron, and Apurv Gupta.

The technology is intended to expand the use of secondary data beyond its traditional roles in recovery and retention. It enables organisations to use secondary data as a knowledge source for generative AI applications without replicating sensitive information into separate AI environments or altering existing data protection systems.

For enterprises, the approach is relevant because large volumes of secondary data are often stored in emails, databases, and virtual machines. The RAG layer is designed to make this data semantically searchable and available to large language models while helping organisations manage data distribution and security considerations.

Within the Cohesity Data Cloud platform, Cohesity Gaia is designed to help organisations use historical data in AI-driven workflows. The platform aims to provide access to insights from stored data while operating within existing security and governance frameworks.