Privacy Policy

1. Core Architectural Tenet: Privacy as Code, Not Policy

At QMatus, privacy is an architectural invariant. We do not rely on after-the-fact legal compliance to protect data. Our systems enforce privacy through strict server-only collections (allow read, write: if false), hardware-level cryptographic isolation, and edge-localized model execution.

2. Trait Sensitivity Gating (72 Dimensions)

In consumer applications such as Oh That's Me (OTM), raw trait vectors are never directly accessible to client devices or third parties. Trait exposure follows strict threshold gating:

• Low Sensitivity: Minimum confidence 0.25, minimum evidence count 2.
• Medium Sensitivity: Minimum confidence 0.35, minimum evidence count 3.
• High Sensitivity: Minimum confidence 0.45, minimum evidence count 4.
• Protected Dimensions (L8 Identity/Values): Explicit, affirmative opt-in confirmation required before any social projection.

3. Zero Data Egress in Enterprise SLMs

Phase 2 and Phase 3 architectures guarantee that client enterprise data never leaves client boundaries. Small Language Models (SLMs) trained by QMatus are deployed on-premises or on private edge hardware, completely eliminating third-party AI vendor egress and multi-tenant data contamination.

4. Inquiries & Verification

For technical security audits or formal data processing verification, contact the founding engineering team directly at privacy@qmatus.com.