Need-to-Know Access Controls and Data Security for LLMs

Knostic is a GenAI guardrails vendor that brings need-to-know access controls to large language models used in enterprise environments. Rather than relying solely on traditional file- or repository-level permissions, Knostic sits between organizational data sources and generative AI interfaces to regulate what information can flow into (and out of) AI conversations based on predefined authorization policies.

It evaluates both incoming user queries and outgoing model responses in real time, so the model’s output stays within each individual’s permitted knowledge boundary as defined by the organization.

A key capability is topic-level segmentation. Knostic can classify corporate knowledge domains by sensitivity and enforce access rules at that level, helping to prevent accidental disclosure of restricted content. For example, financial figures, employee compensation details, merger activity, strategic initiatives, or other confidential corporate assets beyond a user’s clearance.

The platform includes inspection gateways that monitor prompts and generated content across multiple enterprise AI channels — including Microsoft 365 Copilot implementations and custom corporate LLM interfaces — and produces reports that quantify potential data exposure while providing administrators with policy configuration tools.

Knostic was founded by Gadi Evron and Sounil Yu, and it earned notable industry recognition as one of the three finalists in the RSAC Launch Pad 2024 competition.

Knostic was the only startup to win both the RSA 2024 Launch Pad competition and the Black Hat 2024 Startup Spotlight competition. In March 2025, the company announced that it had raised $11 million in a round led by Bright Pixel Capital with participation from Silicon Valley CISO Investments (SVCI), DNX Ventures, Seedcamp, and unnamed angels. Total funds raised now stand at $14.3 million.