Minimize the data surface.

KernelWright Systems LLC designs its inference services to keep customer content separate from the operational metadata needed to secure, measure, support, and bill the service.

Policy ownerKernelWright Systems LLC

Privacy contactprivacy@kernelwright.dev

Policy scopeKernelwright inference services

Data handling

Content should transit the service, not become its dataset.

Kernelwright separates customer content from the small amount of metadata required to operate the service.

Prompts and generated content

Processed to provide the requested inference response; not retained by default

No content retention

Token IDs and embeddings

Not logged or retained as operational telemetry

Excluded

Training use

Customer inputs and outputs are not used to train models

Never used

Operational telemetry

Request shape, timing, token counts, error class, and accounting identifiers

Metadata only

Security events

The minimum information needed to investigate abuse or protect service integrity

Purpose limited

Service providers

Infrastructure and billing providers used only to operate and secure the service

Restricted access

Purpose

Requests are processed to provide inference responses, operate and secure the service, enforce service limits, provide support, maintain accounting records, and comply with law. Kernelwright does not sell customer content.

Telemetry

Operational records may include timing, token counts, request shape, endpoint, response status, error class, and accounting identifiers. They are designed to exclude raw prompts, generated content, token IDs, and embeddings.

Security

Access to operational systems is limited to authorized personnel and service providers with a business need. Security records are used only to investigate abuse, incidents, and threats to service integrity.

Your requests

Questions or requests concerning personal data may be sent to privacy@kernelwright.dev. Security reports should be sent to security@kernelwright.dev.