Security & privacy

Last updated: 2026-06-15

Tola handles documents and data from firms in regulated sectors — mortgage, insurance, accounting, construction and recruitment. Our controls are deliberate rather than decorative. This page describes what we actually do today, in plain English. If something is important to your IT, risk or compliance team and isn't covered here, email hello@tola.ie.

1. Human approvals

Every workflow is scoped around specific human review points. By design:

• No client-facing message is sent without a human approving it in a review queue, logged with reviewer, timestamp and context.
• No regulated advice, lending decision, policy bind, submission or application completion is automated — these actions remain with your regulated people.
• Low-confidence document classifications and ambiguous extractions are routed to a reviewer rather than auto-processed.
• System write-backs (e.g. CRM status updates) are gated on approval rules you define per workflow.

2. Access management

We request the narrowest permissions needed for the specific workflow and nothing more:

• Microsoft 365 and Google Workspace access uses OAuth with scoped application permissions. Your admin reviews and grants the exact scopes needed.
• SharePoint and OneDrive access is granted per site or library, not tenant-wide.
• CRM and practice-system access uses the narrowest API role or token available for the workflow.
• For portals without APIs, authorised browser automation runs under credentials you control and can revoke at any time, where permitted by the portal's terms.
• Access can be revoked by your administrator at any time directly in the connected system.

3. Data processing

• Documents and data are processed in EU regions.
• We do not train models on your data. Prompts, documents and extracted content are not used to improve any provider's foundation models.
• Where we use third-party AI providers to run model inference, we do so under contracts that prohibit training on customer data.
• Where feasible, documents are processed and discarded in-memory rather than persisted. Where they must be persisted (e.g. to support exception review and audit), retention is governed by the policy below.

4. Auditability

Every workflow produces a continuous audit trail:

• Document classifications, extractions and the model's reasoning context at the time.
• Drafts prepared, approved, edited or rejected — and by whom.
• System write-backs: what changed, when, under which approval.
• Access events — which account read which document or made which API call.

Audit logs are exportable on request to support internal audit, compliance review or incident investigation.

5. Document retention

Retention windows are configured per workflow to match your firm's existing policy and the regulatory obligations of your sector. Defaults are set during scoping; you can change them at any time with notice. On termination of an engagement, we delete data on an agreed schedule (default: 30 days after final export).

6. Integration permissions

All integrations run under permissions you grant. You can revoke them at any time:

• Microsoft 365 — Entra ID (Azure AD) admin console.
• Google Workspace — admin console → Apps → Google Workspace Marketplace apps.
• CRMs / practice systems — your platform's API token or app management settings.
• Browser-based automations — credential rotation in the underlying portal.

7. Secure development

• Secrets and credentials are stored in managed secret stores, not in source code or configuration files.
• Infrastructure runs on major cloud providers with current security practices (encryption at rest and in transit).
• Changes to production workflows are reviewed and logged.

8. Reporting a security concern

If you believe you've found a security issue affecting Tola, email hello@tola.ie. We'll acknowledge within one business day and give you a named contact for follow-up.

9. Contact

For any question about this page, your firm's controls or an upcoming compliance review, email hello@tola.ie.