Every AI system in the company, visible and under control.
A lightweight gateway routes AI and connector traffic through one point you control. You see every model, agent and MCP connector in use, decide what each may reach, and keep a complete record without adding work for your team.
What Govern does.
Shadow AI discovery
Finds the models, agents and connectors people already use, including the ones nobody approved, on workstations and servers.
Living inventory
Every AI system with an owner, a purpose, the data it reaches and its risk level. Updated from traffic, not from a spreadsheet.
Gateway-enforced policies
Allow, block or require approval per tool, per system and per data class. Enforced before the call is made, not documented after it.
Threat protection
The AssurAI model inspects calls for prompt injection, data exfiltration and unsafe tool use, and blocks or holds them.
Central deployment
Publish approved skills, prompts, connectors and configurations to every team. Versioned, staged and revocable in one step.
Usage, cost and evidence
Spend per team and model, incident records, and evidence packs mapped to EN 18286 and the AI Act, generated continuously.
From install to evidence in four steps.
Install the gateway
Deploy on your infrastructure or a dedicated EU tenant and connect your identity provider. Typically within the first week.
Discover
Traffic builds the inventory. Within days you see which AI is in use, by whom, and what it reaches.
Set policy
Start from policy packs for the AI Act, GDPR and your data classes, then tune per team. Exceptions route to the right owner.
Run and record
Monitoring agents keep watch, your team decides the few exceptions, and the evidence accumulates for audits and cover.
Built to fit an enterprise estate.
- Deployment
- Self-hosted on Kubernetes or Docker, or a dedicated tenant in an EU data centre.
- AI clients
- OpenAI- and Anthropic-compatible endpoints, MCP clients and servers, locally hosted models.
- Identity
- Single sign-on with Microsoft Entra ID, Okta, Google Workspace and other OIDC providers.
- Data residency
- Prompts, responses and logs stay in the EU. Nothing is shared between customers.
- Audit log
- Tamper-evident record of every model and tool call, exportable to your SIEM.
- Standards mapping
- EN 18286:2026, AI Act Articles 9, 12, 14 and 72, ISO/IEC 42001, NIST AI RMF.
- Open core
- Built on an MIT-licensed open-source control plane your security team can audit. Our own work is the evidence model, risk score, trained model and carrier integration.
Frequently asked.
Will this slow my team down?
No. Approved use passes straight through the gateway. Monitoring agents handle routine traffic, and only the few cases that need judgement reach a person.
Does AssurAI see our data?
Govern runs in your environment or a dedicated EU tenant. Your prompts and logs stay there; AssurAI staff do not access them without your permission.
We already have a DLP or CASB. Do we still need this?
Those tools see network traffic, not what an agent asks a tool to do. Govern adds the AI layer: which model, which connector, which action, and whether policy allows it.
How long does installation take?
Design partners are installed in week one. Larger estates are rolled out team by team.
Works with
See Govern on your estate.
Book a demo on sample data, or join the design partner programme and have Govern installed in week one.
Request a demo