Privacy Policy

1. Introduction

Quantized Labs, Inc. ("Quantized Labs", "we", "our", or "us") respects your privacy. This Privacy Policy explains how we collect, use, disclose, and safeguard your information when you visit our website or use the Quantized Labs and MCaaS platforms.

2. Data Processed During Distillation (MCaaS)

Our core service involves the compression of neural network weight matrices (e.g., .safetensors files). We adhere to a strict Zero-Retention Policy for proprietary customer models:

• Uploaded model weights are processed exclusively in volatile memory (RAM).
• Once the compiled `.quantized` artifact is generated and downloaded, the original source payload is cryptographically wiped from our ingest servers.
• We do not track users across third-party websites or sell personal data to advertisers.

3. Local RAG (Retrieval-Augmented Generation)

By integrating Quantized Labs into your applications, you enable Local RAG architectures. This guarantees that user documents are never uploaded to a cloud provider for vectorization or chunking.

Supported Ingestion Formats (Zero-Cloud)

4. Encryption at Rest & BYOK

All `.quantized` artifacts support hardware-backed AES-256-GCM encryption. Enterprise customers can implement Bring Your Own Key (BYOK) topologies.

• Weights remain encrypted on disk and are only decrypted via Secure Enclave keys at the exact moment of inference.
• If a device is compromised or physically stolen, the local LLM weights and vector databases remain mathematically inaccessible.
• Full integration with AWS KMS, Azure Key Vault, and physical smart cards.