Security & privacy
What base9 reads, where it's kept, who else touches it, and what happens at the end. Plain language, and only what's true today.
base9 reads your helpdesk through an account you control and can switch off. It can't change or close tickets.
Every record belongs to your workspace, and every table and query is scoped to it.
No article changes until someone on your team approves it. Sending it on to your help center is a separate step.
Pilot data is deleted when the pilot ends. A workspace you made yourself is deleted when you ask.
What's read: resolved tickets in the helpdesk project you choose, with their comments, or from a file of tickets you upload. Also your help center (a Zendesk help center, a Confluence space, or a public site read from its address) and any documents you add, one at a time or as a folder. Reading a Google Drive or Dropbox folder you share is rolling out with pilots.
What's stored: not the tickets. Each resolved ticket is read once and summarised: what was asked, what fixed it, the sentence the fix was taken from, and details such as product, dates and a link back to your helpdesk. The conversation, and who the customer is, aren't stored. The summary is checked in code for leftover emails, phone numbers and names. Also stored: your articles and documents, the fixes drafted for them, what you decided and who decided it, and a question from "Try a question" when you choose to save it.
Where: in our database in the US (AWS, US East), in your own workspace. Every record carries your workspace, and every query is limited to it.
How long, in a pilot: for the length of the pilot. At the end it's deleted unless you ask us to keep it. How long anything is kept after that is agreed in writing before we start.
How long, in a workspace you made yourself: until you ask us to delete it. Ask, and it is deleted with everything in it.
Who at base9 sees it: only the people running your pilot or looking after your workspace, to set it up, to fix a problem you report, and to review results with you.
To summarise a resolved ticket, the whole ticket is sent to an AI model over its API, and only the summary is kept. Your articles, and the summaries, are sent the same way to be compared and to draft fixes. base9 runs on Google's Gemini today. Tickets and articles are processed in the US. If a different provider or region is to be used for you, it is agreed in writing first.
We only use the provider's business API. The same checks apply after the AI answers, in code: every quote is matched word for word against the article or the ticket it came from, every sentence of a new article is checked against its sources, nothing changes without a person's approval, and words you have said must never appear in an article are caught before a fix can be approved.
Instructions written inside a ticket are treated as data, never as instructions to the AI.
The complete list of companies that process data for base9:
| Company | What for | Where |
|---|---|---|
| Vercel | Hosting the app and this website | US |
| Supabase (on AWS) | Database and file storage | US East |
| Google (Gemini API) | Reading and summarising tickets, comparing them with your articles, drafting fixes | US |
| Resend | Emails about your account: approval, password reset, invitations, folder approval | US |
| Formspree | This website's demo request form | US |
Systems you connect yourself (your helpdesk, your help center, a Confluence space, a shared-drive folder) are where your data already lives. base9 reads from them with the access you give; they are not subprocessors of base9.
base9 Triage is optional and set up on request, so none of this applies unless you have asked for it. It reads new tickets as well as resolved ones, and it stores more: the tickets it has read, each decision it made with the exact articles and context it used, what your agents did with each draft, and the customer's organization details. A reply reaches a customer only when one of your agents approves it. Internal documents guide its decisions but are never quoted or named to a customer, a rule enforced in code.
No analytics, no tracking pixels, no advertising tags and no cookies. We can't tell who visited which page.
The only thing collected is what you type into the demo request form: your work email and company website, plus your name and helpdesk if you choose to give them. Formspree forwards it to us by email; Vercel keeps standard server logs, including IP addresses, for a limited period; unpkg serves one script the form uses. We use it only to reply to you, never for a mailing list.
Ask us what we hold about you or your company and we'll tell you. Ask us to delete it and we will, without asking why. If you're in the UK or EU, the GDPR rights of access, correction, erasure, portability and objection apply, and you can complain to your data protection authority.
base9 operates from Kathmandu, Nepal. The providers above store data outside your country, so data is transferred internationally on that basis. Ask us for the specifics before sharing anything.
Questions, or a security review or vendor questionnaire before you start: contact@base9.io.
Last updated 11 October 2026. If anything here changes, this page changes the same week.