Data & video usage policy
LYNX is designed for on-device inference. This page describes how the SDK handles your data — what stays local, what (if anything) leaves your environment, and under what conditions.
This document is provided for technical and procurement review. It describes the SDK's data handling behavior and is not a substitute for your license agreement, which governs the commercial terms of use.
Core principle: inference is local
LYNX runs inference entirely on your device. Your images and video are processed locally and are never transmitted to Synetic or any third party as part of normal inference operation.
- Image and video input stays on your hardware
- Detection, segmentation, tracking, and all other inference runs on-device
- No frames, no detections, and no derived data are sent anywhere during inference
What requires network access
The SDK contacts Synetic infrastructure only in these specific cases:
| Operation | When | What is sent |
|---|---|---|
| License activation | One-time, first run | License key, machine registration data |
| Model download | First use of a model | License validation; model is downloaded to your device |
| Model updates | On online startup, if a newer model exists | License validation; updated model downloaded |
| Trial verification | Periodically during a trial | License validation only |
| Telemetry | When online, unless opted out | Operational usage metrics — not inference inputs |
| Feedback submission | Only if you explicitly call submit_feedback() | The image you choose to submit |
None of these involve transmitting your inference inputs automatically. Your images and video stay on your device. After activation and model download, inference itself operates offline.
Feedback submission is opt-in
The SDK includes a submit_feedback() function that lets you send an
image to Synetic to help improve future model training. This is
entirely optional and only sends data when you explicitly call it. The
SDK never submits your data automatically.
Telemetry
The SDK records operational telemetry and surfaces it in your dashboard when an internet connection is available. This covers operational metrics about SDK usage — not your inference inputs. Your images and video are never transmitted as part of telemetry.
Telemetry can be disabled by setting the opt-out flag in the SDK.
Local diagnostic logging (e.g. LYNX_INFERENCE_LOG) writes to your
local stderr only and is separate from dashboard telemetry.
Model weights
Model weights are encrypted and tied to your license. They are downloaded to and cached on your device. The weights themselves contain no customer data.
Data residency and offline deployment
Because inference runs locally and your video never leaves your hardware, LYNX is suited to:
- Privacy-sensitive deployments where video cannot leave the premises
- Regulated environments with data residency requirements
- Disconnected environments (telemetry and trial verification require periodic connectivity; paid licenses can be configured for extended offline operation)
For air-gapped deployment requirements, contact sales@synetic.ai to discuss activation and offline licensing options.
Questions
For security review, data processing agreements, or compliance documentation, contact sales@synetic.ai.