Privacy Policy.

If you do not want to read the full legal text, understand our three core engineering principles:

1. We do not train models on your data. Your proprietary routing trees, customer calls, and knowledge base are used strictly for inference (running the application).
2. We do not sell your data. We route voice traffic; we are not data brokers.
3. We enforce strict isolation. Your workspace is cryptographically separated from all other tenants on the Syntax network.

Syntax Voice ingests three categories of data to provision and operate its routing infrastructure on your behalf:

• Account Telemetry. Billing information, dashboard IP logs, and login credentials.
• Communications Media. Real-time SIP / WebRTC audio streams, SMS logs, and transcribed call text generated during an Inbound Session.
• Workspace Logic. Custom knowledge bases, webhook API keys, and routing trees you upload to Syntax Flow.

Syntax Voice utilizes Large Language Models (LLMs) to power the Syntax Core Intelligence cognitive engine. We operate under a strict zero-data-retention policy with our LLM providers. When an audio transcript is sent via API for intent recognition, the provider is legally and programmatically barred from storing that payload or using it to train their foundational models. The data exists in memory only long enough to return the inference, and is then destroyed.

When a call escalates to a Syntax Specialist, that US-based operator is granted temporary, restricted access to the live session transcript and the specific routing protocols required to resolve the ticket. All Specialist endpoints operate on zero-trust architecture. Access is programmatically revoked the millisecond the ticket is closed, and Specialists cannot download or export audio media.

Call logs, transcripts, and audio recordings are retained based on the custom TTL (Time-To-Live) parameters set by the Workspace Admin — defaulting to 30 days for standard accounts. If an Admin deletes a workspace or initiates a data destruction request, all associated media, telemetry, and webhooks are permanently purged from our active servers within 72 hours, and from encrypted backups within 30 days.

To deliver sub-800ms latency globally, we partner with industry-leading infrastructure providers. A current list of our sub-processors (e.g., AWS, Cloudflare, Twilio, Stripe) is available upon request. We ensure all sub-processors are bound by Data Processing Agreements (DPAs) that meet or exceed our internal standards. See our DPA for the legal framework and sub-processor change-notice policy.