Privacy Policy
Last updated: June 2026
TorCyte is built on a simple principle: your data stays yours. This page explains what we collect, what we don't collect, and what happens to information your device shares with our servers.
1. Camera and microphone — on-device processing
If you choose to enable sensor learning (optional), your browser may access your camera or microphone. Raw video frames and audio recordings never leave your device.
Here's exactly what happens:
Visual (camera)
Your camera captures a frame. Before anything is transmitted, your browser shrinks it to a 32×32 pixel thumbnail and converts it into 64 numbers that describe general shapes, brightness patterns, and structure — similar to describing a painting as "warm colours, soft edges, bright centre" rather than describing the actual image. These numbers carry no detail about faces, text, or anything identifiable. Only these 64 numbers (roughly 512 bytes) are sent to our servers.
Audio (microphone)
Your microphone picks up sound. Before anything is transmitted, your browser analyses which frequencies are present — bass, mid, treble — using the same perceptual scale your ears use. The result is 64 numbers describing the tonal character of the sound: "low rumble", "bright sharp tone", "quiet mids" — not words, not a recording, not a voice fingerprint. Only these 64 numbers are sent to our servers.
We also apply a novelty gate: if the scene hasn't changed meaningfully since the last transmission, nothing is sent at all. This further reduces the volume of data leaving your device.
2. What we store
| Data type | Stored? | Details |
|---|---|---|
| Account (email, name) | Yes | Required for authentication and billing. |
| API call logs | Yes, 90 days | Endpoint, plan, latency, status code. No request body content. |
| Concept embeddings (64 floats) | Yes, in engine | Abstract numeric vectors. Not reversible to original image or audio. |
| Raw camera frames | Never | Processed and discarded on your device before transmission. |
| Audio recordings | Never | Processed and discarded on your device before transmission. |
| Concept labels (text) | Only if you provide one | Optional human-readable label you attach to a submission. |
3. Shared Brain (opt-in)
The Shared Brain is a collective learning network. If you opt in, your concept embeddings (the 64-float vectors described above) are aggregated with those of other opted-in users to improve the shared base model. The aggregation is anonymous — your identity is not attached to the concept vectors in the shared pool.
- Opt in — enable via the dashboard toggle. Your concept vectors contribute to the next shared model update.
- Opt out — disable the toggle at any time. Contribution stops immediately; the shared model is not rebuilt retroactively, but your vectors no longer influence future updates.
- Delete my contribution — request full removal via Settings → Privacy → Delete my shared brain data. The shared model is rebuilt from remaining opted-in users without your data. An audit record of the deletion (timestamp only, no content) is retained for compliance.
4. Data retention and deletion
You can delete your account at any time from Settings → Account. This removes your authentication record, API keys, subscription state, and call logs. Engine concept data associated with your user ID is also purged within 30 days.
5. Third parties
- Stripe — payment processing. Card details are handled entirely by Stripe; we never see or store them.
- No analytics scripts. We do not use Google Analytics, Meta Pixel, or equivalent third-party tracking.
- No ad targeting. Your data is never sold or used for advertising.
6. Contact
For privacy questions or data requests, email privacy@aivantor.com. We aim to respond within 5 business days.