How BumbleB uses your data
This page explains, in plain language, what happens to your data when you use BumbleB. The binding versions of these commitments are in our Terms of Service, Privacy Policy and Data Processing Agreement. They say the same thing, in the same words.
Our three promises
No training. BumbleB will not use Customer Content to train, retrain or fine-tune any AI model, and will not allow any sub-processor to do so. The only exception is a Submitted Conversation, which a user chooses to send us.
Usage Data. BumbleB records how the Service runs — which tools were called, in what order, and whether they worked — without names, account identifiers or the values in your data. We use Usage Data only to operate, secure, measure and improve the Service.
No human reading. BumbleB staff do not read Customer Content to improve the Service. Staff read it only to provide support you ask for, to investigate security or abuse, or when you send us a conversation.
These apply to every customer: individuals and companies, on every plan, in every country.
Three kinds of data
| Customer Content | Account Data | Usage Data | |
|---|---|---|---|
| What it is | Your questions, files and connected data; what our tag collects from your website visitors; and every answer, chart and table we produce for you. | Who you are and how you use your account: names, emails, logins, plan, invoices, support tickets. | Our record of how BumbleB ran — which tools it called, in what order, and whether they worked — with no names, account IDs or data values. |
| Who owns it | You | Us | Us |
| What we use it for | Answering your questions; keeping the Service secure; support you ask for | Running and billing your account; security; telling you about your account and the product | Running and improving BumbleB |
| Used to train AI models? | Never — unless you send us a conversation yourself | No | Only our own internal components, such as the part that decides which tool to use |
What we record when BumbleB answers a question
BumbleB answers by calling a series of tools — filter the data, compare two periods, draw a chart. For each call we keep a short record like this:
| You ask | What we keep in Usage Data |
|---|---|
| “Why did trial-to-paid conversion drop in March?” | Question type: period-comparison · Step 1: segment · Step 2: filter(<category>) · Step 3: compare(<metric>, <period>) · Step 3 failed, retried, succeeded · Answer accepted |
We keep the kind of question, not your words. We keep that a filter ran, not what you filtered on. No names, no account, no numbers from your data.
How we tell whether BumbleB got it right
- Live labelling. While answering, BumbleB tags each question with a type from a fixed list,
such as
period-comparison. We keep the tag, not your words. - Automatic signals. Did a tool fail? Did BumbleB retry? Did you rephrase, or accept the answer?
- Automated checking. A separate model compares your question with the steps BumbleB took and records a verdict such as “right tool, wrong time period”. It stores the verdict only.
- Re-checking recent sessions. For up to 90 days, automated tools may re-check recent sessions for a new kind of problem. They don’t look at data about your website visitors, and they output labels only. No person reads the sessions. Individual users in India are asked first, with a box that starts unticked.
- Submitted Conversations. If you press 👎 or “Send to BumbleB”, you’re choosing to share that conversation. Our team may read it and use it to improve BumbleB, including to train our models. Data about your website visitors is hidden before anyone reads it. You can withdraw it at any time; we then delete it, though a model already trained with it isn’t retrained. Company admins can turn this off for their whole workspace; it’s off by default on Enterprise plans.
Who can read your content
| Who | When |
|---|---|
| You and your team | Always |
| BumbleB support | When you ask for help with a specific issue. They look at it in place and don’t copy it anywhere. |
| BumbleB security | To investigate abuse or a security incident |
| BumbleB product team | Only conversations you’ve chosen to share |
| Our sales and customer-success team | Never. They see which features your account uses and how often — never what you asked or what the answers said. |
What we never do
- Train any AI model on your content (unless you send us a conversation).
- Let our AI providers train on your content.
- Use data about your website visitors to improve BumbleB.
- Build profiles of individual people.
- Sell or share your data.
- Link Usage Data back to you.
How long we keep things
| Data | How long |
|---|---|
| Your conversations — questions and answers | 90 days, then deleted; backup copies within a further 90 days |
| Records of who accessed what (no content) | 12 months |
| Submitted Conversations | 12 months, or until you withdraw them |
| Account Data | While your account is open, then as long as the law requires |
| Usage Data | As long as it’s useful. It identifies no one, so it isn’t deleted with your account. |
| Prompts held by our AI providers | OpenAI: up to 30 days for abuse monitoring under its terms. Through OpenRouter: not stored (zero retention), apart from brief retention a provider must keep by law. Details on our sub-processor list. |
We keep data longer only when the law requires it, for example during a legal dispute.
Your choices
- Delete your content at any time, or ask us to.
- Send us or withdraw a conversation.
- Get a copy of your data, correct it, or have it erased — see our Privacy Policy.
- India: contact our Grievance Officer, Chinmaya, at grievance@bumbleb.co. We respond within 90 days.
Questions
Email privacy@bumbleb.co. For contracts and data-processing terms, email legal@bumbleb.co.