Cold-Serve LabsPrivacy policy
This policy covers the inference service run by Cold-Serve Labs ("Cold Serve", "we").
Logging and training at a glance
| Question | Answer |
|---|---|
| Are prompts or inputs stored after the request? | No |
| Are outputs stored after the request? | No |
| Is request or response content written to logs, crash reports or disk? | No |
| Is your data used to train or tune models? | No |
| Is data sold? | No |
| What is kept? | Request metadata (time, model, token counts, latency, status code) for up to [N] days |
Zero data retention
We do not store the content of requests or responses (prompts, outputs, images, tool calls) after a request finishes. Content is not written to logs, crash reports or disk, and is never used to train or tune models.
What we keep
- Per request: time, model, token counts, latency and status code. We use these for billing, rate limits and uptime.
- For API customers: account and billing contact details.
We keep request metadata for up to [N] days, and billing records as long as tax law requires.
Sharing
We do not sell data. We share billing data only with our payment processor, and data only when the law requires it.
Where it runs
Requests are processed on our own servers in Japan.
Operator
[Operator name and address, as required by Japan's Act on the Protection of Personal Information]. Questions: privacy@coldservelabs.com.