Privacy at Boundless

This page summarises how Boundless handles the data you send it. The binding documents are the Inference Services Customer Terms, where Section 3 covers data, and the Boundless Networks privacy policy, which applies to every Boundless product. Where this page and those documents differ, those documents govern.

Zero data retention#

Your prompts and outputs are never stored and never used to train a model. This service does not write request or response bodies to a log or a database.

The claim covers the content of your requests. Each request still leaves a billing record, described below.

What is recorded#

For each request the service keeps what is needed to bill it and to let you check the bill:

• The model that served it
• Input, cached-input, and output token counts
• A timestamp
• The computed cost
• The API key and team the request was billed to

The usage page shows you exactly these fields, and they are enough to recompute any charge from the rate card yourself. None of them contains the text of a prompt or a completion.

Billing records are retained for 90 days, then deleted automatically.

Account data#

Signing in creates an account record: your email address, your name if you provide one, your team membership and role, and your API key names. Authentication is handled by a third-party identity provider. Payment, where card payment is enabled, is handled by Stripe, and card numbers never reach this service.

Third-party infrastructure#

Models run on compute operated by third parties. Section 3 of the terms sets out the limits on what can be promised about infrastructure that is not ours. Read it if zero retention is a compliance requirement for you, and write to us if you need those limits in writing for a specific deployment.

Questions, and requests about your data#

Email support@boundless.network for anything on this page, including access, correction, and deletion requests. See Support.