GoRefer Trust Center
AI Security & Governance
Updated April 2026
GoRefer's AI assistant, Gio, operates under a strict data governance framework designed specifically for the tax industry. This page explains exactly what data Gio can and cannot access, how AI interactions are isolated per tenant, and the controls that prevent data leakage.
We understand that using AI in a tax workflow requires a higher bar of trust. The details below are written for both technical reviewers and non-technical firm owners.
Your data is never used to train AI models
GoRefer makes a firm commitment: no customer data — including any input to Gio — is ever used to train, fine-tune, or improve any AI model. Your clients' information only flows in one direction: into your workspace. It never flows into model weights or shared learning systems.
What Is Gio?
Updated April 2026
Gio is GoRefer's built-in AI assistant. It can draft communications, surface referral insights, summarize client activity, suggest commission structures, and help preparers manage their workload. Gio is deeply integrated with firm data — which is why its data access is tightly controlled.
What Gio Can Do
Draft client-facing emails and referral communications
Summarize referral pipeline and commission trends
Answer questions about firm data (scoped to your firm only)
Suggest workflow improvements based on aggregated activity
Assist with document drafting and compliance checklists
All actions are fully audited and reversible
What Gio Cannot Do
Access raw SSNs, EINs, bank details, or driver's license numbers
Read data belonging to a different firm
Retain memory or context across sessions (stateless per request)
Autonomously save or modify client records without human approval
Share outputs with third parties
Initiate outbound communication on your behalf without explicit approval
What Data Is Sent to AI Infrastructure?
Updated April 2026
The table below documents each data category that Gio may encounter and the protection measures applied before any data leaves GoRefer's own infrastructure.
| Data Type | Sent to AI? | How It's Protected | Sensitivity |
|---|---|---|---|
Client SSN / EIN / Bank details | Never | PII fields are decrypted only in tightly scoped service calls; the AI layer never receives raw PII | Critical |
Client name | Optional — anonymized | Gio workflows can receive pseudonymous references; full names are excluded by default unless explicitly included by the firm | Controlled |
Referral workflow data | Yes — in structured context | Tenant-scoped. All AI context is associated with a single firm and never mixed with other firms | Standard |
Commission calculations | Yes — aggregated figures | Submitted as structured data without personally identifying the preparer unless firm has configured otherwise | Standard |
Analytics & usage patterns | Yes — aggregated only | No individual-level data; firm-level aggregate metrics used for insights | Low |
AI Security Controls
Updated April 2026
Every AI interaction on the GoRefer platform is governed by the following technical and policy controls. All controls listed below are actively enforced — not aspirational.
| Control | Description | Status |
|---|---|---|
Tenant isolation | Every AI request is scoped to a single tenant. Context from one firm can never appear in another firm's AI responses. | Enforced |
No model training on customer data | Customer data submitted to Gio is never used to train, fine-tune, or improve any AI model — including GoRefer's own systems. | Enforced |
PII stripping before AI calls | An automated pre-processing layer strips or masks PII fields before any data is dispatched to AI infrastructure. | Enforced |
AI response audit log | Every Gio interaction is logged with actor, timestamp, input summary (not full content), and response metadata. Stored for 365 days. | Enforced |
AI disclaimer on all outputs | Every AI-generated response includes a mandatory disclosure that the content was AI-generated and should be reviewed before use. | Enforced |
Human review before saving | AI outputs are presented as drafts. No Gio response is saved to a client record without explicit acceptance by a firm user. | Enforced |
Fallback infrastructure | If the primary AI provider is unavailable, Gio fails gracefully to a secondary provider. Customer data is not retained by either provider beyond the request lifecycle. | Enforced |
Prompt injection detection | Input sanitization and output validation layers are in place to detect and block prompt injection attempts that could cause unintended data exposure. | Enforced |
AI Credit System as a Governance Mechanism
Updated April 2026
GoRefer's AI credit system is not just a billing mechanism — it serves as a natural rate-limiting and auditability layer. Every Gio operation consumes credits, creating a one-to-one log of every AI action performed on your data.
Rate Limiting
Credit consumption prevents unbounded AI data access
Unusual spikes in credit usage trigger Security Hub alerts
Admins can set per-user AI credit limits
Full Auditability
Every credit deduction maps to a specific AI action and actor
Credit logs are immutable and exportable
365-day credit history retained by default
Access Control
Credit allocations are scoped per user role
Admins can revoke AI access for individual users instantly
AI features follow the same RBAC as the rest of the platform
AI Infrastructure Isolation
Updated April 2026
AI processing uses zero-retention endpoints
GoRefer's AI infrastructure is configured to use zero data retention agreements with our AI service providers. Data submitted for AI processing is not stored, logged, or used for model improvement by those providers after the request completes.
Request Lifecycle
Context assembled in GoRefer's own infrastructure
PII fields are stripped/masked before the context leaves GoRefer
Single-request dispatch
AI calls are stateless — no session state is maintained between requests
Response validated before delivery
AI outputs are scanned for unexpected PII or policy violations before returning to the user
No cross-tenant context
The context window for each request is fully scoped to the requesting firm
Failure & Fallback Handling
Primary / fallback provider architecture
If the primary AI provider is unavailable, requests route automatically to a backup provider with the same data governance requirements
Graceful degradation
If both providers are unavailable, Gio fails with a clear user-facing message — no data is exposed or cached during the failure
Incident logging
All AI provider failures are logged in GoRefer's internal monitoring system with full request metadata (no data content)
Prompt Injection Protection
Updated April 2026
Prompt injection is a class of attack where malicious input attempts to override AI instructions, extract data from the AI context, or cause unintended behavior. GoRefer applies multiple layers of defense against these attacks.
Input Defense
User inputs are sanitized before inclusion in AI prompts
Known injection patterns are blocked at the input validation layer
Client-submitted text is structurally segregated from system instructions
Character limits enforced on all AI input fields
Output Defense
AI responses are validated against expected output schema before delivery
Responses containing unexpected PII patterns are flagged and blocked
Output sanitization prevents injection artifacts being rendered to other users
Annual penetration testing includes AI-specific prompt injection test cases