Book a meeting

AI policy that acts.

Fourgovia sits between the prompt and the internet, and rules on every prompt before it reaches an AI model. Try it.

AI assistant assistant.example/chat
You

Draft the Acme renewal email to maria.silva@acme.com: value US$ 1.2M, tax ID 51-2446789.

Fourgovia

REDACTED 3 sensitive items masked before sending.

AI received
Draft the Acme renewal email to [EMAIL]: value [AMOUNT], tax ID [TAX ID].

Entered in the audit log · EVT-0417

Policy
Examples

Demo · runs in your browser · nothing you type is sent

Book a meetingHow it works
01 RISK02 GOVERNANCE03 COMPLIANCE04 AUDITEMAILREDACTED[EMAIL] ← maria.silv…AMOUNTREDACTED[AMOUNT] ← US$ 1.2MTAX IDREDACTED[TAX ID] ← 51-2446789TEXTPASSED22 tokens · clean

Fig. 1 · EVT-0417 · Sales Public chatbot · 22 tokens · 3 sensitive · Redacted

  • Sensitive
  • Leaves
  • Absorbed

Your AI policy is on the wall. Your data is in the prompt.

Policies, training and acknowledgement checkboxes all depend on someone remembering, at the exact moment they paste. Bans push AI use out of sight. Either way the data leaves, and leadership answers for it.

Policy 4.2 · Use of generative AI

Employees must not enter confidential information, customer data or source code into generative AI tools.

  • Approved by the board
  • Acknowledged by all staff
  • Reviewed annually
  1. 14:32Sales Public chatbot

    Summarize this contract for Acme Corp, renewal US$ 1.2M and draft the email.

  2. 14:41Support Internal agent

    Refund João Pereira, card 4111 1111 1111 1111, order #88213.

  3. 15:07Engineering Code assistant

    Fix this config: AKIAIOSFODNN7EXAMPLE fails on deploy.

Fig. 2 — The prompt, the same afternoon. Illustrative scene.

A rule that can be ignored is not a control.

Four pillars. One path every prompt takes.

The name says it: four pillars, and one via, one path, through them. Every prompt crosses the same four layers, in order, before anything reaches an AI model.

01020304
Fig. 3 — The four layers a prompt crosses, from the inside out.
  1. 01

    RiskDetect

    The prompt is examined in transit for the data that must not leave.

  2. 02

    GovernanceDecide

    Your policy is applied to that prompt at the moment it is sent.

  3. 03

    ComplianceAct

    The ruling is carried out: blocked, redacted, or the employee is warned.

  4. 04

    AuditRecord

    Every event is entered in the record, as evidence for regulators and the board.

Four rulings, each with its own signature.

  • BlockAbsorbed

    The prompt stops before it leaves. Nothing reaches the model.

  • RedactMasked

    The sensitive part is stripped out. The rest goes through.

  • WarnHeld

    The employee is told why, in the moment, before anything leaves.

  • LogRecorded

    The prompt goes through and the event is entered for audit.

Fig. 4 — How each ruling reads in the event display.

At the gateway, in the API, on the device.

Three ways to deploy Fourgovia, covering the ways your company actually uses AI.

Your people and systems

  • Employees at work
  • Engineers in their tools
  • Apps and agents you build

Fourgovia

  • Network gatewayInspects outbound AI traffic at the network layer.
  • API / SDK layerSits in front of the LLM APIs your own apps and agents call.
  • Endpoint agentRuns on company devices.

AI

  • Public AI chatbotsThe assistants employees open in a browser.
  • Internal AI apps and agentsBuilt by your teams on LLM APIs.
  • Code assistantsThe tools your engineers write code with.
Fig. 5 — Prompts carrying sensitive data go in; only what your policy allows comes out.

Built for the people who answer for it.

RoleAnswers forWhat Fourgovia gives them
Board & C-levelLiability and reputationProof that AI policy is enforced, not assumed, in a form you can put in front of regulators and the board.
Compliance & riskControls that holdControls that enforce themselves at the point of use, with the evidence trail already written.
Legal & DPOData protectionPersonal and confidential data stopped or masked before it leaves, and a record of every event.

The first question you will ask us.

Does Fourgovia see our prompts? Here is what we can state today. The rest is being finalised; ask us in your meeting.

What Fourgovia inspects

Content pending · Fourgovia to confirm

What Fourgovia stores

Content pending · Fourgovia to confirm

Where it runs

As a network gateway, an API / SDK layer in front of LLM APIs, or an endpoint agent on company devices.

Content pending · Fourgovia to confirmHosting options (SaaS, on-premise, hybrid): pending.

Data residency and retention

Content pending · Fourgovia to confirm

Encryption and access control

Content pending · Fourgovia to confirm

The demo on this page

Runs entirely in your browser. Nothing you type is sent anywhere.

Ask us in your meeting
  1. A policy no one follows is a liability, not a control.
  2. Training fades. Checkboxes get clicked. Bans get bypassed.
  3. Enforcement happens at the prompt, or it doesn’t happen.
  4. We don’t write the rules on the wall. We carry them out.

The people behind it.

Content pending · founder details

See Fourgovia act on your data.

Book a meeting. We will show you Fourgovia on your own use cases.

What happens next

  1. We reply to your work email.
  2. We agree a time that suits you.
  3. We show Fourgovia on your use cases.

We use these details only to arrange and prepare your meeting. See our Privacy policy.