August 5, 2024 at 01:15PM
Public generative AI tools have gained immense popularity, leading to new chat capabilities and features by many applications. However, connecting these large language models to internal knowledge repositories poses risks of oversharing sensitive information. Knostic addresses this with per-user need-to-know control and aims to expand its solutions to various software-as-a-service tools. Knostic competes in the Black Hat Startup Spotlight competition and offers a technology showcase at their booth.
From the meeting notes, it’s clear that the intense popularity of public generative AI tools has raised concerns about potential risks associated with connecting large language models (LLMs) to internal knowledge repositories. This risk includes oversharing sensitive information due to inferences made by LLMs, potentially exposing organizations to increased risk.
Knostic, a company focused on addressing this issue, defines per-user need-to-know and creates a knowledge control layer. Their technology aims to provide organizations with visibility, control, and curation. They continuously query genAI tools, such as Microsoft’s Copilot, to identify unexpected exposures and modify permissions for content to prevent oversharing. They also aim to address the need-to-know problem beyond Copilot and Glean, extending to all software-as-a-service tools incorporating LLMs.
Knostic is a finalist in this year’s Black Hat Startup Spotlight competition and plans to demonstrate its technology at the event.
Additionally, if the company were a band, its name would be “Guardians of Gnosis” (thrash metal), and its mascot would be a barn owl symbolizing knowledge and wisdom.