Privacy Policy & Context Invariants
Effective Date: May 01, 2026 • Version 2.1 (Sovereign Data Governance)
Zero Foundation Model Retraining Guarantee
KlecherAI does not use private user queries, source code, uploaded files, or chat histories to train proprietary foundation models. Context memory is stored in AES-GCM-256 encrypted relational partitions on Neon PostgreSQL and strictly scoped to your explicit tenant ID.
1. Context Document Isolation
We differentiate between Context Documents (persistent artifacts like syllabi, resume, codebase docs) and Chat Attachments (ephemeral files). Neither document class is shared cross-context without explicit user grant.
2. Independent Sharing vs. Training Consent
Generating a public sharing link or export never grants model training rights. Public link snapshots exclude context documents and private memory facts by default.
3. Telemetry & Minimization
Product telemetry is pseudonymous and limited to latency, token metrics, error codes, and plan tiers. Raw user prompts, secrets, and repository payloads are strictly excluded from logs.
4. Data Deletion & Export Rights
You have the permanent right to export your complete context history, purge memories, or delete your entire account. Deletions cascade immediately to remove chunks, vectors, and relational metadata.