VERIFIED
How we protect your data
- Data is stored encrypted at rest.
- TLS is enforced for all data in transit.
- Tenant isolation is enforced via Supabase Row Level Security policies.
- Your company data is never used to train AI models.
VERIFIED
Who can see what
- Knowledge access is restricted by role, department, authority level, confidentiality classification, and Visibility Rules.
- Even administrators cannot view flows outside their authorisation scope in full.
- A five-layer confidentiality matrix governs field-level visibility across all operations.
VERIFIED
How operations requiring approval are paused
- Financial thresholds are defined in configuration.
- Critical operations never auto-proceed; they route to the appropriate person for Human Approval.
- Operations awaiting approval are visible and traceable in the Restricted Action queue.
VERIFIED
How the Audit Trail works
- All significant decisions and policy enforcements are written to an immutable audit log.
- The log operates on an append-only model; historical records cannot be modified.
- A DELETE=false RLS policy prevents audit records from being deleted.
VERIFIED
Multi-model orchestration
- Model selection is managed within the platform according to task, quality, security, cost, and governance requirements.
- There is no dependency on a single model provider.
- Open-source LLMs are never used. All models are enterprise-grade with contractual data protection terms.
VERIFIED
Platform vs. company administrator
- Platform management and tenant management are strictly separated.
- Orygent platform staff cannot read your company's confidential content, documents, meeting minutes, reports, attachments, or financial and personal data.
- Support access is limited to the operational, log, and technical data needed to run the service, and is governed by recorded break-glass sessions.
VERIFIED
Data deletion and portability
- Deletion requests are processed under GDPR Article 17.
- Data portability is provided under GDPR Article 20.
- Requests can be submitted to privacy@orygent.com.
VERIFIED
Trust Control Layer
- A dedicated policy engine sits between every AI action and execution.
- Five rule layers, role, department, authority, confidentiality, and channel, must all clear before any operation proceeds.
- Any rule violation automatically triggers a RESTRICTED state and routes to the Human Approval queue.
- The Trust Control Layer cannot be bypassed by any user, including platform administrators.