Lasso Security
End-to-end security for generative AI and agents

Need-to-know access controls for enterprise AI.
Knostic is an independent, seed-stage startup (roughly $19M disclosed as of August 2026) addressing a real and narrow problem: LLM oversharing in tools like Copilot and Glean. Its founders are well-regarded in security, and it has won credible industry competitions, which lends legitimacy. That said, it is early-stage, its category is nascent, and its value depends heavily on which AI assistants you deploy. Pricing is sales-led with no public list, so budget expectations should be set through a direct conversation.
Knostic is a GenAI knowledge-security platform that prevents enterprise AI assistants like Microsoft 365 Copilot and Glean from oversharing sensitive data. It simulates LLM responses across user access profiles, detects oversharing, and can block or remediate answers so they follow need-to-know rather than can-access logic. Founded in 2023 by well-known security veterans and backed by around $19M in disclosed funding, it is an independent but early-stage vendor in a nascent category.
Knostic tackles a specific and increasingly urgent enterprise problem: LLM-powered assistants such as Microsoft 365 Copilot and Glean can synthesize and surface information a user technically has permission to reach but shouldn't necessarily see, creating an oversharing and data-leakage risk that traditional permission and DLP tools miss. Knostic's platform simulates and tests how these assistants respond across different user access profiles, detects oversharing patterns, and can dynamically block or remediate responses so answers follow need-to-know logic aligned to organizational roles. The company frames this as a readiness and governance layer for safe enterprise AI adoption. It integrates with the AI search and assistant tools organizations are deploying, provides monitoring and prompt-simulation capabilities, and surfaces where knowledge boundaries are being crossed. This positions Knostic alongside data-governance tooling (such as Microsoft Purview) as a complement that catches inference-level exposure those tools don't address. Knostic was founded in 2023 by cybersecurity veterans Gadi Evron and Sounil Yu and is based in Herndon, Virginia. It is independent and venture-backed: it raised an $11M seed round led by Bright Pixel Capital in March 2025 and later added a $5M SAFE investment as an RSAC 2025 Innovation Sandbox Top 10 finalist, bringing disclosed funding to roughly $19M. It won notable startup competitions (RSA Conference Launch Pad and Black Hat Startup Spotlight in 2024), but remains an early-stage vendor in a fast-forming category.
Knostic is a GenAI knowledge-security platform that prevents enterprise AI assistants from oversharing sensitive data. It simulates and tests LLM responses across user access profiles and enforces need-to-know controls in tools like Microsoft 365 Copilot and Glean. Founded in 2023 by security veterans Gadi Evron and Sounil Yu, it has raised roughly $19M and remains independent. It is credible but early-stage in a nascent category.
Knostic was founded in 2023 by Gadi Evron and Sounil Yu and is headquartered in Herndon, Virginia. It focuses on securing enterprise adoption of LLM-based assistants by preventing oversharing and enforcing need-to-know access.
The company is independent and venture-backed. It raised an $11M seed round led by Bright Pixel Capital (with SVCI, DNX Ventures, Seedcamp and angels) in March 2025, and added a $5M SAFE as an RSAC 2025 Innovation Sandbox Top 10 finalist, bringing disclosed funding to roughly $19M.
Knostic simulates how AI assistants answer questions across different user access profiles, detecting cases where the assistant would surface information beyond a user's need-to-know. It analyzes responses against user context and access policy and can dynamically block or remediate risky answers in real time.
It integrates with tools such as Microsoft 365 Copilot and Glean and positions itself alongside data-governance tooling like Microsoft Purview, catching inference-level exposure that static access and DLP controls miss. Prompt simulation, response analysis and continuous monitoring are core capabilities.
Mid-to-large enterprises and regulated organizations deploying LLM assistants such as Copilot and Glean, where data-leakage and oversharing risk is a barrier to adoption.
Security analysts and AI-governance teams who run simulations and manage oversharing findings.
CISOs and heads of data security responsible for safe AI rollout.
Microsoft 365 administrators, compliance officers and data-governance leads.
A mid-to-large enterprise, often in a regulated sector, actively deploying Microsoft 365 Copilot or Glean and needing to prove that AI answers respect need-to-know boundaries.
Approximately $19M in disclosed funding: an $11M seed round led by Bright Pixel Capital (March 2025) plus a $5M SAFE tied to the RSAC 2025 Innovation Sandbox. Reported totals vary slightly across sources.
It stops enterprise AI assistants like Microsoft 365 Copilot and Glean from oversharing sensitive information by enforcing need-to-know controls on the answers they generate.
No. It is positioned as a complement that catches inference-level oversharing that traditional access and DLP tooling can miss.
Yes. As of August 2026 Knostic remains an independent, venture-backed company with roughly $19M in disclosed funding.
It was founded in 2023 by cybersecurity veterans Gadi Evron and Sounil Yu, and is based in Herndon, Virginia.
No. It uses an enterprise sales model, so you need to contact the vendor for a quote.
Side-by-side pages for pricing, features, and best-fit use cases.
End-to-end security for generative AI and agents
Security platform for AI models and the ML lifecycle
10Web is an AI-powered WordPress website builder and hosting platform that generates complete websites from user prompts
Copy.ai is a GTM (Go-To-Market) AI platform designed for sales, marketing, and revenue teams to automate workflows including sales outreach, content creation, lead processing, and ABM campaigns.