Privacy Policy
We do not ask you to trust a vague privacy promise. We give you a client-side architecture with a narrow, inspectable data boundary.
File content never uploaded
The files you process stay in browser memory on your device. We do not receive the contents of your contracts, images, invoices, or recordings.
Your device is the AI server
Model inference happens in your browser through a WebGPU-capable architecture with local fallbacks when needed. The core workflow does not depend on sending the file to our backend.
Clear boundary, minimal logging
We do not build hidden profiles from file contents. Like most hosted sites, infrastructure logs may still record basic access metadata needed for operations and abuse prevention.
1. Introduction
Effective Date: January 1, 2026
PrivaKit is designed for people who need AI help with sensitive files but do not want those files routed through a cloud processing pipeline. This policy explains the technical boundary behind that claim.
2. Data Processing Model
PrivaKit uses a client-side architecture. In practical terms, that means:
- Image Processing (Background Removal, Upscaling): When you choose an image, it is decoded inside the browser and processed with local CPU/GPU resources. The resulting file is generated on-device unless you explicitly export it yourself.
- Audio Transcription: Audio files are decoded and transcribed locally in the browser. We do not need the recording content on our servers to provide the feature.
3. Information Collection
Because of this architecture, we do not receive or store the following categories of content during normal use:
- The content of files you upload (photos, audio, video)
- Processed results
- Your camera or microphone real-time data (unless you explicitly authorize it locally for recording, and that data is not uploaded)
3.1 Automatically Collected Non-Personal Information
Like most websites, our hosting and delivery providers may record limited request metadata so the site can stay available and secure. This may include:
- IP address (only for security protection and DDoS mitigation)
- Browser type and version
- Access time and date
These logs describe a visit to the site, not the contents of files you process. They exist for hosting, security, and aggregate operational visibility.
4. Cookies and Local Storage
We may use browser LocalStorage to save your preference settings (such as: whether AI model cache has been downloaded, interface language preference). This data is only stored on your device and is not sent to us.
To optimize model loading speed, the browser uses caching mechanisms to store AI model files (such as .onnx files). This is a standard browser feature designed to save your bandwidth and waiting time.
5. Third-Party Services
We may rely on infrastructure providers such as hosting, CDN, and model delivery services. Those requests can expose standard web metadata like IP address, browser version, and request time, but they do not give us the actual file contents processed inside your browser session.
6. Children's Privacy
Although our services are general and do not collect data, we do not knowingly market to children under 13.
7. Policy Changes
As technology evolves, we may update this Privacy Policy. Any significant changes will be prominently notified on this page. Given that our core "no data collection" architecture is unlikely to change, the core commitments of this policy will remain unchanged.
8. Contact Us
Since we don't collect your contact information, if you have any questions or suggestions about this Privacy Policy, please contact us through public channels provided in our community or code repository.