Trust3 AI Integrates Agent DOS with Microsoft Agent 365 to Deliver Real-Time Agent Security Across the Enterprise
Trust3 AI today announced a deep integration between Agent DOS and Microsoft Agent 365, extending enterprise agent
Press Release Disclaimer: This is a press release distributed through the XPR Media network. It has not been independently verified by our newsroom.
![]()
Trust3 AI today announced a deep integration between Agent DOS and Microsoft Agent 365, extending enterprise agent security to every AI agent operating across the Microsoft ecosystem. As organizations accelerate AI agent adoption across Microsoft Copilot Studio, Microsoft Foundry, Microsoft Fabric, and OneLake, the attack surface is expanding faster than traditional security controls can follow. The integration connects Microsoft’s agent registry and identity infrastructure to Trust3 AI’s real-time enforcement engine, so agents are discovered, assessed, and secured before they act.
Microsoft Agent 365 provides a centralized registry in the Microsoft 365 admin center that records each agent’s owner, models, tools, MCP servers, connected agents, and, Entra Agent ID, and it aggregates risk signals from Microsoft Purview, Microsoft Entra, and Microsoft Defender. Agent DOS uses that record to enforce what each agent may access and do, on Microsoft platforms and beyond, and stops unauthorized actions as they happen.
Three Capabilities That Change How Enterprises Secure Agents
The integration delivers three capabilities organizations need to run AI agents at scale without sacrificing security or control:
- Discover: Every agent registered in Agent 365, with its owner, model, tools, MCP servers, connected agents, and Entra Agent ID where assigned, joins the unified Agent DOS inventory alongside agents already detected on Databricks, Snowflake, and other platforms. Shadow agents and unmanaged agents are flagged on first appearance. No agent goes unaccounted for.
- Observe: Agent DOS traces the tool calls Agent 365 records to the systems, data sources, and tables each agent touches, in Microsoft Fabric, OneLake, and beyond. Security teams see what each agent is doing, what it reaches, and whether that matches its declared purpose.
- Secure: Microsoft’s aggregated risk signals from Entra, Purview, and Defender feed directly into the Agent DOS Trust Score. Agent DOS evaluates each agent against the sensitivity of the data it reaches, using the organization’s existing classification, scopes which tools it may call, and carries verified user identity across every hop from agent to tool to data. When a threat is detected, a kill switch stops the agent at the platform level, the gateway, or both.
Protecting AI ROI Across Microsoft Fabric and OneLake
As enterprises build AI-powered workflows on Microsoft Fabric and route data through OneLake shortcuts from Databricks and Snowflake, a critical security gap emerges: the data moves, but the access policy doesn’t. OneLake shortcuts expose tables from source platforms without carrying the row filters, column masks, or grants that govern them at the source, leaving Fabric Data Agents and Copilot to query data under OneLake’s permissions alone, not the organization’s intended security model.
Agent DOS extends the security boundary that governs human access to agents, across OneLake, Fabric, and every connected source. Trust3 AI’s Permissions Sync translates Unity Catalog and Snowflake policies into native OneLake Security roles in near-real time, so every agent operating in Fabric inherits the same access decision as the source platform intended. Policy stays consistent across the data estate, with no re-authored controls and no proxy in the query path.
That makes agent security a question of AI ROI as well as risk. Agents that can’t be trusted with sensitive data can’t be deployed for high-value use cases. Agents that reach beyond their authorization create liability. Trust3 AI gives the confidence to enterprises to deploy agents at scale in Microsoft Fabric, OneLake, Copilot Studio, Foundry, and hybrid environments, because every access decision is governed, auditable, and enforced.
“Every agent in your environment is a potential attack surface. Microsoft Agent 365 gives you a registry of your agents, their owners, and Microsoft’s risk signals. Agent DOS enforces what each one may reach, so no agent goes beyond what it’s explicitly authorized to do. And with OneLake and Fabric in the picture, that boundary has to extend all the way to the data. That’s how enterprises move fast with AI without losing control.”
— Neeraj Sabharwal, Co-Founder, Trust3 AI
Why Agent Security Is Now a Board-Level Priority
Organizations now face a reality where AI agents outnumber human users in many enterprise environments, operate persistently across systems, and make decisions that carry real business and compliance consequences. Without a security control plane, agent sprawl creates invisible risk: compromised agents that exfiltrate data, misconfigured agents that breach data boundaries, and orphaned agents that remain active long after their owners leave.
Agent DOS and Microsoft Agent 365 together provide the foundation enterprises need:
- Unified agent inventory across Microsoft and non-Microsoft platforms
- Identity-aware access enforcement tied to Microsoft Entra Agent ID
- Risk scoring grounded in signals from Microsoft Purview, Entra, and Defender
- Real-time tool gating that restricts what agents can call and when
- Policy sync that extends source-platform access controls into Microsoft Fabric and OneLake
- A kill switch that stops any agent at the platform, the gateway, or both
This is not observability. It is enforcement — and it is what responsible AI deployment at enterprise scale requires.
About Trust3 AI
Trust3 AI is the security control plane for enterprise AI agents and the data they reach. Built by the creators of Apache Ranger, which protects data in many of the world’s largest cloud and Hadoop deployments, Trust3 AI extends that proven governance model to the age of autonomous AI agents.
Agent DOS is Trust3 AI’s agent security platform. It discovers every agent running across frameworks and clouds, scores each one through a real-time Trust Score engine, and enforces least-privilege access at the moment of action, before data moves or a tool call fires. Agent DOS is vendor-neutral and framework-agnostic, with native integrations across Microsoft Copilot Studio, Microsoft Foundry, Microsoft Fabric, OneLake, Databricks, Snowflake, AWS Bedrock, Salesforce Agentforce, and more than 50 additional data sources and SaaS platforms.
Trust3 AI is a Microsoft Partner, NVIDIA Inception member, Snowflake Startup Accelerator participant, Databricks Built On Partner, AWS Qualified Software vendor, and Google Cloud Select Technology Partner.
Learn more at trust3.ai or request a live demo at trust3.ai/demo.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261117228509/en/
Media gallery

