Training Data Rules
This document defines the strict boundaries and protocols governing how data is utilized to refine Pixizen’s intelligence. We prioritize data minimization and user sovereignty above all else.
1. Scope & Eligible Sources
Pixizen uses training data exclusively to improve AI accuracy, reduce bias, and maintain system safety. We draw from two primary categories:
- Licensed & Public Data: Open-source datasets and properly licensed third-party libraries.
- Consent-Based User Data: Prompts and files submitted by users who have explicitly opted in.
Training data is never utilized for advertising, behavioral profiling, or third-party resale.
3. Explicit Consent Requirement
Pixizen enforces a Strict Opt-In mechanism for all AI training activities:
- Default State: Participation is disabled by default for all accounts.
- Active Grant: Users must manually toggle training permissions.
- Revocability: Consent can be withdrawn at any time with immediate effect on future pipelines.
4. Strict Prohibitions (Non-Negotiable)
The following data types are never ingested into training pipelines:
- Personal Identifiable Info (PII)
- Financial & Billing Data
- Medical & Health Records
- Government Identifiers
- Private Communications
- Biometric Identifiers
5. Anonymization & Security Isolation
De-Identification: Before training, data undergoes automated tokenization and aggregation. Anonymized data is processed separately from live production environments.
Environment Isolation: Training occurs in segregated, access-restricted "Air-Gapped" compute environments. This data is never accessible to other customers or third-party vendors.
8. Data Lifecycle & User Rights
We apply the Data Minimization Principle: irrelevant or redundant information is purged immediately. Users retain the full right to access, delete, or restrict the processing of any data they have contributed to the training pool.
10. Audit & Transparency
Zenbitx LLC maintains detailed training logs, consent records, and internal compliance audits to ensure accountability as the Data Controller. No third party is permitted to reuse or re-extract Pixizen training data.