Knostic’s AgentMesh Joins VirusTotal's Crowdsourced AI Program
8 June 2026
Use CasesZero-Trust controls aren’t bulletproof. See whether your employees have the ability to infer sensitive insights from LLM responses.
The Scenario
You’ve rolled out least-privilege groups, created network segments, and granted conditional access. But if someone asks Copilot:
“Summarise current R&D cost-cutting plans and their HR impact.”
Copilot’s seemingly-harmless answers risk exposing strategies you can’t find in one file. Security needs proof that Zero-Trust also holds at the knowledge layer, not just the network layer.
Create copies of interns, contractors, or power users without impacting production.
Probe finance, HR, IP, and exec topics through Copilot/Glean.
Knostic flags risky answers and maps how they happen across employee boundaries.
Sort and filter by department, role, sensitivity level, or user context.
One click tightens labels and shores up your LLM to end oversharing.
Confirm whether segmentation truly blocks inference of sensitive information.
Highlight low-privilege users who can still piece together sensitive insights.
Make it easy to track and manage risk levels over time.
Apply fixes and prove success without waiting for the next audit.
Combined, these capabilities prove (or disprove) your Zero-Trust posture at the speed of AI.
Core Capabilities
Connects to M365 and AI tools in minutes, so you can get started faster.
Analyzes the information your LLM can access to uncover how sensitive information can be inferred.
Validates your adjustments and ensures the leaks you’ve blocked stay blocked.
Identifies each failed Zero-Trust boundary, fixes it, and can retest to ensure the leak is sealed.
Stores a tamper-proof log of prompts, sources, and risks.
