Skip to content
Legal

Legal

No-training policy

We do not train models on your prompts or outputs. This policy explains exactly how request and response content is retained, the 30-day operational window that applies across the platform, the limited metadata we keep to bill and to stop abuse, how data is handled when it passes through an upstream provider, and how to request deletion.

Last updated: June 23, 2026. This No-Training Policy is incorporated into the Terms of Service and works alongside the Privacy Notice. “Customer content” means the prompts, files, and other inputs you send and the responses the models generate for you.

We do not train on customer data

AdvancedMind AI does not use your customer content to train, fine-tune, or otherwise improve any model, ours or our providers'. This applies to every account and every route, with no setting to enable. We use customer content only to serve the request you made and to operate the platform within the retention rules below.

Retention

Every model declares a retention mode that governs what happens to request and response content after a call completes. Every model in production today runs under the same mode, so you can confirm how a call will be handled before you run it.

30_day_ops

Our models run under 30_day_ops. Request and response content may be retained for up to 30 days for operational purposes (debugging, reliability, and abuse investigation) and is then deleted. Content under this mode is never used to train models.

The applicable mode is set by the model, not chosen per request, and is shown on the models page and on every request in Usage. For how retention interacts with Research Mode, see the Research Mode Policy.

Metadata we keep

Even after content is deleted at the end of its retention window, we keep a small amount of metadata about each request. This metadata does not include the content of your prompts or outputs. It includes:

  • Billing and metering: the model used, input and output token counts, the computed cost, the API key prefix, the organization, and the timestamp. This is what lets us debit the correct amount and show you accurate usage.
  • Abuse and safety: request status, the research tier and the retention mode the model declares, and signals needed to detect and stop policy violations. This is how clear abuse is blocked and high-risk access is reviewed.

We retain this metadata for as long as needed to bill, to meet legal and tax obligations, and to investigate abuse. It is never used to train models. You can see your own metering metadata in Usage.

Provider pass-through

AdvancedMind AI routes requests to upstream model providers to serve inference. When your request is served through a provider, the relevant content passes through that provider to generate the response. We contract with providers so that customer content sent through our routes is not used to train provider models, consistent with the retention mode in force. Providers may process content transiently to return a response and may keep limited operational logs under their own terms. We select and configure routes to honor the retention commitment declared for each model. Provider routing details are recorded internally for reliability and accounting; they do not change the no-training commitment.

How to request deletion

You can ask us to delete customer content and associated records, subject to the limits below:

  • What we delete on request. Retained request and response content under 30_day_ops, ahead of its normal 30-day deletion window.
  • What we may retain. Billing, tax, and abuse-related metadata that we are required to keep to operate the service, meet legal obligations, and enforce the Acceptable Use Policy. This metadata does not include your prompt or output content.
  • How to ask. Send a deletion request from the account contact on file to [email protected], identifying the organization and the scope of the request. We confirm when deletion is complete.

For the full account of what data we collect and your rights over it, see the Privacy Notice.


Summary

  • We do not train on customer content, on any route.
  • 30_day_ops retains content up to 30 days for operations, then it is deleted. It applies to every model in production.
  • We keep limited metadata (model, tokens, cost, key prefix, status) to bill and to stop abuse, never to train.
  • Provider pass-through is configured to honor the no-training commitment and the model's retention mode.
  • Request deletion at [email protected]; required billing and abuse metadata may be retained.

Related