Does AI store or learn from my data?

We use Anthropic (Claude) as our primary AI provider. Anthropic states that it does not train its models on data sent through their API.

Your conversations are processed in real time to generate responses and stored in your DIVORSAY account. Anthropic's stated policy is to retain the text sent to and from the model for up to 30 days under its standard API terms, so its trust and safety team can review for misuse, and not to use it for model training.

Anthropic also states that content flagged by its automated trust-and-safety systems may be retained for up to two years. We can't tell you in advance what gets flagged — that's their system, not ours — so we'd rather you know the possibility exists than assume a shorter window applies to everything.

If you would rather a sensitive detail not go to an AI provider at all, you can leave it out of what you type to Auntia. Documents you upload to the Vault are also read by an AI provider in order to organize them, so the same applies to what you upload.