Should an AI browser expose provider credentials in the app?

No. A browser app shipped to users should not expose reusable AI service credentials in client-side code. A backend can hold credentials server-side and enforce quotas.

What context does an AI browser need?

Only the context needed for the request: the prompt, recent chat turns, current page text, selected text, mentioned tabs, and attached images when relevant.

AI Browser Privacy

An AI browser should send the minimum useful context for the task, keep ordinary browsing data local, avoid client-side provider credentials, and clearly explain when cloud processing happens.

The Core Privacy Problem

A browser assistant is powerful because it can see the page you are on, understand selected text, read tabs you mention, and take actions. That same context is sensitive. Good design means the assistant should not send everything all the time.

What Should Stay Local?

Normal browsing history and bookmarks.
Passwords and passkeys.
Generated image files after they are returned.
Settings, pins, and most browser state.

What Might Need Cloud Processing?

Cloud AI needs the prompt and relevant context. For Breeze, Nav requests can include current-page text, @-mentioned tabs, recent chat turns, attached image data for edits, and a client identifier for daily quota.

Why Not Put Provider Credentials in the Browser?

Any reusable credential bundled into a shipped client app can be extracted eventually. Breeze Cloud keeps provider credentials server-side, applies daily limits, and forwards requests without storing those credentials in the app.

What Breeze Does

Breeze uses Nav through Breeze Cloud for chat, browser actions, image generation, and image editing. The app keeps the browser experience local where it can and uses cloud processing only when you invoke Nav. The Privacy Policy explains that Breeze Cloud may connect to supported AI service providers as features evolve.