Privacy Architecture
We built SwiftSay with a "Privacy by Design" philosophy. Here is exactly how your data flows and who has access to it.
Audio & Content Processing
To ensure zero-latency and maximum security, your audio data is transmitted via an encrypted tunnel directly from your device to our Secure GPU Processing Nodes.
Direct Pipeline
The SwiftSay management server acts only as an arbitrator—negotiating security tokens and managing your account. It never sits in the middle of your audio stream, ensuring the lowest possible latency.
Physical Isolation
Your audio packets never touch the SwiftSay management servers where your database and admin account details reside. They are routed exclusively to a specialized processing cluster.
Data Visibility Matrix
| Data Type | Visible to Admins? | Purpose |
|---|---|---|
| Email & Account ID | YES | Account management and technical support. |
| Usage Stats (Words/Minutes) | YES | Billing analytics and system load monitoring. |
| Raw Audio Files | NEVER | Processed in transient memory; never touches our database servers. |
| Transcribed Text Content | NEVER | Handled via direct pipeline between GPU nodes and your client. |
| Credit Card Details | NEVER | Securely tokenized and handled by Stripe. |
Payment Security
We take financial security seriously. SwiftSay utilizes Stripe, the world's most trusted payment processor, to handle all subscriptions and transactions.
- PCI Compliance: All credit card details are tokenized and processed directly on Stripe's secure infrastructure.
- No Local Storage: SwiftSay servers never see, handle, or store your raw credit card number, CVV, or expiry date.
- Encrypted Webhooks: Our systems only receive a unique Customer ID and a status update (e.g. "Paid") via cryptographically signed webhooks.
Infrastructure & Sub-Processors
SwiftSay utilizes high-performance GPU compute providers to host the SwiftSay Pro Model. These providers are contractually bound to strict data privacy standards and are prohibited from training models on user data or retaining audio packets beyond the duration of the transcription request.