If you use Bridge, you're going to put real ideas into it. Product ideas. Feature sketches. Half-formed thoughts that might turn into a company. It's fair to want to know what happens to those ideas once you press record.
This post is the plain-English version of the privacy policy. It describes what Bridge actually does with your data, in the order things happen.
What gets captured
When you record a note or type one into Bridge, three things happen:
- The raw note — text, or a transcribed voice memo — is stored on Bridge's cloud infrastructure under your account. This is the sync layer; it's what lets you open the app on another device and see the same notes.
- A minimal amount of metadata — timestamp, device type, how long the audio was — is stored alongside it. No location, no contact data, no device identifiers beyond the ones needed for the account.
- Nothing else. Bridge does not access your contacts, your calendar, your photo library, or the other notes in your phone's Notes app. The surface of what Bridge sees is exactly what you put into Bridge.
Audio recordings are transcribed to text and then discarded unless you explicitly choose to retain them. By default, you end up with the transcript and we don't keep the voice file.
What goes to the LLM
When you run the Interview phase, Consolidation, or Specification, the relevant notes are sent to an LLM provider — Claude (Anthropic), GPT-4 (OpenAI), or Gemini (Google), depending on what you've chosen. That provider processes the content and returns a response.
A few things to know about this step:
- The content is sent through the provider's API, not the consumer chat product. API usage has stricter data-retention rules than chat usage. None of the three providers train on API content by default.
- Only the notes relevant to the current operation are sent. When you run an interview on one brief, the other briefs in your account are not transmitted.
- You can see exactly what was sent. Each AI operation logs the input tokens and output tokens associated with it.
If you're working on something where even API-level transmission to a third-party LLM is not acceptable — some regulated contexts, some competitive-sensitive domains — Bridge is not the right tool for that content today. A local-model option is on the roadmap but is not part of the MVP.
What Bridge does not do
A few things that are easy to assume and are not true:
- Bridge does not sell your data. There is no advertising model, no data broker relationship, no anonymized-analytics-resale pipeline. The business model is subscriptions.
- Bridge does not use your content to train any model, its own or anyone else's. The API providers don't either, at the tier Bridge uses.
- Bridge employees do not read your notes as a routine practice. Access to stored content is gated and logged, and the only times it's accessed by a human are for explicit support requests you initiate.
- Bridge does not share your content with other users. There is no "public feed," no "discover what others are planning," no ambient social layer. Your briefs are yours.
What you can do
- Export. You can export every note and every generated artifact (schema.sql, openapi.yaml, component-tree.md, plan.md) at any time. The export is a plain ZIP of files. You own the output.
- Delete. Deleting a brief removes it from Bridge's storage. Deleting your account removes your content entirely, on a schedule documented in the privacy policy.
- Choose your LLM provider. You pick which provider handles your AI operations. If you prefer Anthropic's data handling policies to OpenAI's, or vice versa, that choice is yours.
The honest caveat
Bridge is a pre-release product from a small company. We are not a Fortune 500 vendor with a dedicated compliance team. We have written the privacy policy in plain language, we will not hide material changes, and if we ever have to change something important we will tell you directly rather than through a mandatory-acceptance modal.
If you have a specific question about how Bridge handles a specific kind of content, email us. Real answer, real person, reasonable timeframe.
That's the whole data story. Keep it in mind, decide for yourself whether it's a fit, and use the tool with open eyes.