Trust3 builds Trust3 AI, a security and governance platform for enterprise AI agents and the data those agents can reach. It is aimed at security and data teams that need to know which agents are running, what they are allowed to do and what they actually did.
The problem is that agents are being deployed faster than anyone can track them, often with broad access to data they do not need. Trust3 gives organisations one place to govern them.
At A Glance
| Vendor | Trust3 |
|---|---|
| Category | AI agent security and governance |
| Identities covered | AI agents, agent service principals, delegated users, MCP servers |
| Deployment | Multi-cloud control plane behind existing AI gateways |
| Agentic AI | Purpose-based access control, MCP Security, A2A Security, Trustscore |
| Market classification | AI Governance & Compliance; also Agentic AI Identity |
What Trust3 Does
Enterprises now run agents on many platforms at once, from Databricks and Azure AI Foundry to Copilot Studio and Microsoft 365, and each agent can touch sensitive data. Trust3 brings the governance of those agents and of the data behind them into a single layer. The approach is least privilege enforced at the moment of each request, before data moves or a tool call runs.
Access decisions weigh the agent’s service principal, its declared purpose and the classification of the resource together. Data rules are then applied natively at sources such as Snowflake, Databricks and BigQuery, so the same policy holds whichever agent asks. For regulated organisations the platform maps controls to the EU AI Act, HIPAA and NIST and produces evidence packs for EU AI Act compliance.
Products
Agent Discovery
Agent Discovery answers the first question most security teams have: which agents exist. It scans frameworks and clouds and raises a flag when an unregistered agent first appears, so shadow AI is caught early rather than found during an incident.
Agent Observability
This module records what each agent does, from prompt to response, in an immutable audit trail. It looks for PII leakage, scope drift, policy violations and other anomalies, and spots drift and prompt injection in real time.
Agent Security
Agent Security is where access is decided. Each request is checked against the purpose the agent declared at registration, so an agent built for one job cannot quietly do another. When an agent needs more, scope can be widened just in time, with every expansion logged and limited in duration. A kill switch pauses or quarantines any agent and revokes its credentials at once.
MCP Security And A2A Security
MCP Security treats every MCP server as untrusted by default and checks it against a live registry tied to each agent. A2A Security verifies identity along agent-to-agent delegation chains, which protects against impersonation and confused-deputy attacks when one agent acts on behalf of another.
Data Security And Trustscore
Data Security gives row, column and tag-based access control at the data source itself. Trustscore grades each agent from 0 to 10 on Policy Compliance, Scope Adherence, Security Posture and Behavioral Baseline, giving teams a simple way to compare agents and decide where to look first.
How It Works
Trust3 runs as a real-time enforcement plane across any cloud, with zero standing access. It does not replace AI gateways such as Portkey or Kong; it sits behind them and adds identity and purpose to each decision.
| Component | What it does |
|---|---|
| Trust3 AI Secure Agent | The enforcement engine of the control plane, which applies policy to each agent action. |
| MCP enforcement | Every tool description is inspected before it enters the agent’s context, hidden directives are removed, and each agent is held to an allowlist of approved MCP servers. A server is blocked if its certificate, host or behaviour changes. |
| Token exchange | Agents never handle raw credentials. The platform issues scoped, short-lived tokens through secure token exchange, tied to the agent’s declared purpose. |
| Governance Intelligence Agent | GIA is the console that brings the discover, observe and secure functions together for administrators. |
Identities And Systems Covered
| AI agents | Each agent registers with a stated purpose, with identity mapping for ephemeral and delegated agents |
|---|---|
| Delegated users | User identity, declared purpose and policy verdicts travel with each agent-to-agent hop |
| MCP servers | Tracked in a live registry per agent |
| Cloud IAM | Over-privileged AWS IAM roles can be replaced with attribute-based access synced to the corporate IdP |
| Agent platforms | Databricks Agent Bricks, Microsoft Copilot Studio, ServiceNow Now Assist, Azure AI Foundry, Claude Code |
| Data platforms | Snowflake, Databricks, BigQuery, Apache Iceberg, Microsoft OneLake and Fabric |
Integrations
| Area | Integrations |
|---|---|
| Agent frameworks and models | LangChain, CrewAI, Microsoft Copilot, OpenAI, Anthropic, Cursor, NVIDIA, Amazon Bedrock |
| Data catalogues and engines | Collibra, Dremio, Starburst |
| AI gateways | Portkey, LiteLLM, Kong, Cloudflare, Apigee |
| SaaS data sources | Salesforce, Slack, ServiceNow, Workday |
Company Background
| Headquarters | Newark, California, United States |
|---|---|
| Leadership | Balaji Ganesan (co-founder and CEO), Don Bosco Durai (co-founder and CTO), Neeraj Sabharwal (co-founder), Anish Menon (Chief AI Security Officer) |
| Funding | $63.5 million in total, from Insight Partners, Sapphire Ventures, Battery Ventures and Accel |
| History | Built by the creators of Apache Ranger |
Note: This overview is compiled from Trust3’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.