Data security overview
Data in transit
All data is transmitted over TLS-encrypted HTTPS connections. API requests are authenticated via Bearer tokens (API keys) validated on every call.
Data at rest — Ephemeral processing
Neuralk operates on a stateless, process-and-discard model:
Training and test data (feature matrices, labels)
Held in memory only for the duration of the inferencerequest. Once the prediction is returned, all data arrays are released from memory. No customer data is writtento disk or persisted in any database.
The API acts as an authenticated proxy:
data is forwarded to the inference engine, processed, and the resultsare returned in the same request cycle.
What we store
We store only operational metadata for billing, monitoring, and debugging purposes:
Stored
Request ID, Timestamp
Model name, latency
Credits consumed
HTTP Status Code
IP Address, user agent
Not stored
Training data (X_train, y-Train)
Test data (X_test)
Predictions / Probabilities
Raw request / response
Feature values or labels
See our Terms & Conditions for public API use.
Authentication & access control
API keys are hashed (SHA-256) before storage — raw keys are never persisted
// A. 01
Optional IP allowlisting per API key
// A. 02
Optional endpoint restrictions per API key
// A. 03
Daily request quotas and rate limiting enforced per key
// A. 04
Organization-level expiration and suspension controls Infrastructure
// A. 05
Hosted on Scaleway Kapsule, region fr-par (Paris, France, EU)
// A. 06
Internal service communication over private Kubernetes networking
// A. 07
No data leaves the EU infrastructure boundary during processing
// A. 08