AI Guides › Getting Started
By Nigel Guy · 2 min read
Most people either assume the worst — everything you type is being read by a person somewhere and sold on — or assume the best — it's private because it feels private, like a notes app. Neither assumption is reliable, and most people never actually check, because privacy policies are written to be skimmed past, not read.
The rule: three specific questions answer almost everything that matters, and each one has a findable answer — you don't have to guess or assume.
| Question | Where to look | Why it matters |
|---|---|---|
| Is my input used to train the model? | The privacy policy or a settings toggle, usually named something like "improve the model" | This determines whether your content could ever influence outputs shown to other people |
| Can I turn that off, and is it on by default? | Account or privacy settings, not the marketing page | Default-on versus opt-in tells you what happens if you never touch a setting |
| How long is it kept, and can I delete it? | Data retention section of the privacy policy, or a data-controls settings page | Determines whether "I deleted the chat" actually means the data is gone |
Assume anything you type could, in principle, be seen by the company running the tool, for support, safety, or abuse-prevention purposes, regardless of the training-data setting — that's a different question from training, and most providers reserve some version of this access. Treat sensitive categories — health details, legal matters, anything you wouldn't put in a work email — with the same caution you'd use with any other online service that stores your input, not with extra trust just because it's a chat interface that feels private.
Skip reading the full legal text of a privacy policy end to end; the three questions above are usually answerable from a summary, a settings page, or a targeted search within the document. Skip assuming a paid tier automatically means stricter data handling than a free one — check the specific claim rather than inferring it from the price.