Data Collection Policy

What AsterWise may collect

When Allow data collection is enabled for your account, AsterWise may retain the request content your tool sends, including supplied conversation history, instructions, tool calls and results, together with model responses, routing information and submitted playground originals.

Drafts and unused uploads are excluded. Text-only playground conversations do not require a stored original object. A successful inference request does not establish that your coding task succeeded.

Purpose and retention

Research records are used for routing research, evaluation and possible future model-training preparation. AsterWise is not training a model with the archive at launch. Research material may be retained indefinitely under this launch policy.

Research storage is best effort. An unavailable archive may produce missing or incomplete records without delaying inference or changing your bill.

Your account preference

Allow data collection is enabled by default for a new account. You can turn it off in Settings. The change applies to future AsterWise research capture for API and playground text and attachments. It does not delete earlier captures and does not stop required operational, security, usage or financial records.

Turning collection off also cannot change content already sent to a model provider or the provider's handling of a later request that is needed to answer you.

Secrets and original files

Text research copies use high-confidence credential redaction, but detection is not exhaustive. Avoid supplying unnecessary passwords, private keys or other secrets.

Original files included in an actual playground provider submission can be copied unredacted and labelled as originals. Research-copy redaction does not alter the content sent for inference.

Model providers

AsterWise sends required request content to the selected model provider. Providers process data under their own applicable terms. The aster-work baseline uses Meta Muse Spark 1.3 Contributor at launch; Meta may use prompts and responses processed on that tier to train future Meta models. This is separate from the AsterWise research preference.