Privacy Policy

Information we handle

• Account information. Google sign-in and Firebase Authentication provide an account identifier and basic profile information, such as your name, email address and available profile picture. WeTalk Money does not receive your Google password.
• Information you share. Imported document text, financial entries, dates, descriptions, amounts, account or debt details, budget plans, filenames and import metadata. Uploads are processed into text and financial records; the app does not provide an archive of the original uploaded binary files.
• Saved conversations. When you allow session storage, the app saves your questions, replies, source references and available response diagnostics so you can review or export a conversation. On a new conversation or reconnection, earlier questions and replies are archived as a compressed transcript, and only a compact memory gist is carried into the new conversation. You can review, edit or clear that gist in Sources & imports. When cloud processing is allowed, the selected reasoning provider may receive the earlier conversation once to prepare its summary; otherwise a conservative local memory is used. A summary can be incomplete or inaccurate and does not create financial records or authorize actions. Spoken questions submitted through live voice are included in that saved conversation.
• Preferences and service information. Voice settings and browser sign-in state, request counts used for rate limits, timestamps and technical information needed to deliver and protect the service. Hosting and authentication providers may process IP addresses, browser information and operational logs.

WeTalk Money currently has no direct bank connection or Google Drive integration. It does not load personal financial records until you share them.

Readable credit-report text can supply dated account balances, limits, payments and statuses alongside the report passages. The importer omits recognized personal-information sections and identifiers, but this is not a guarantee that every report layout removes all personal data. Review what you share. Reports may overlap and do not establish live balances or a complete financial picture. Google Document AI is the default PDF reader. Only after you approve that file, PDF pages are sent through our authenticated backend to Google’s US Document AI processor for paid text recognition. The app does not archive the uploaded original PDF or page groups. Google states that synchronous processing is performed in memory and that customer documents are not used to train Document AI; operational metadata and Google Cloud service terms still apply. See Document AI data handling. You can turn on processing on this device instead: that backup uses self-hosted local text extraction and English OCR without uploading PDF pages. In either mode, you review and may edit extracted text before import; only confirmed text and extraction coverage are retained with your workspace. Google PDF processing approval is separate from conversation and voice permissions. Cancelling stops queued uploads but processing already sent may still finish and be charged. OCR may misread names, amounts, or dates. Cloud intelligence receives confirmed workspace content only under your existing provider consent.

How information is used and shared

We use information to authenticate you, maintain your workspace, detect and map imports, calculate financial summaries, answer questions from your records, provide the voice features you choose, handle support requests and protect the service from misuse.

The hosted app uses Google Firebase and Google Cloud for authentication, hosting, backend processing and stored workspace data. Workspace access is tied to your authenticated account. Authorized service operators may access information as needed to operate the service, investigate security issues or respond to your support request. Information may also be disclosed when required by law.

We do not sell your financial documents or use them for advertising. The current app does not automatically send your analysis exports to a suggestion box or use your saved sessions to train a model. Financial summaries use the app’s calculations. With your cloud-processing permission, your selected reasoning provider (Google Vertex by default, or OpenAI when enabled in Settings) receives a financial workspace overview, current app state and conversation context, plus normalized records and document text or passages needed for your request. Small datasets can be included in full; larger datasets can be queried across every current import. Optional OpenAI services also provide speech and transcription. The app does not send original uploaded binary files to the reasoning model.

See Firebase’s privacy and security information and Google’s Privacy Policy for their handling of information. Providers may process information in the United States and other countries where their services operate.

Cloud reasoning providers

Google Vertex and optional OpenAI reasoning receive the same CAF workspace context and typed query results. Cloud permission covers the selected provider; existing conversations must review the updated provider disclosure before cloud reasoning resumes. Google Cloud’s service terms and retention policies apply to Vertex processing. We do not promise zero retention. The app’s backend authenticates to Vertex using its existing service identity, without a downloadable service-account key.

Optional voice processing

• Browser voices and dictation. Google or other browser service voices may send the text being read to their speech provider. Browser dictation may send microphone audio to the browser’s speech provider. Dictation requires your permission; you can review its transcript before sending. If you keep “Remember my browser dictation choice on this device” enabled, your explicit browser-dictation choice is saved locally across visits. This never starts the microphone automatically or enables OpenAI/live microphone processing. You can turn remembering off or revoke dictation in Settings.
• Built-in speech. The built-in voice is generated by our backend without calling an external speech provider. Luma’s bundled giggle clips play without a speech-provider request.
• OpenAI. Optional OpenAI speech sends the reply text you ask it to read. Cloud transcription sends the recorded audio for transcription. Live voice sends microphone audio when you enable it and sends text for spoken replies. These replies can contain financial information from your workspace. Separate processing permissions are required.
• Gemini Live. Its current settings save a voice preference only. Gemini Live processing is not connected.

The app does not persist raw microphone recordings. OpenAI’s separate processing and retention rules still apply to information sent to it. Its API data is not used for model training by default unless the API customer opts in. Reasoning requests use store:false, which does not override the provider’s retention rules. Some endpoints retain abuse-monitoring data for up to 30 days, with exceptions. We do not promise zero retention. See OpenAI’s API data controls.

The OpenAI reasoning switch starts off and is separate from voice-provider switches. Only the selected reasoning provider is called; the app does not automatically send each question to both. Showing a provider’s voices in Settings is separate from allowing cloud processing. Turning off a permission stops future use of that option; it does not erase information already sent to a provider.

Storage, retention and security

The hosted app stores financial context, snapshots and saved sessions in Firestore. Earlier versions of imported information can remain in snapshots after you replace an import. Saved workspace records currently have no scheduled automatic expiry; they remain until deleted through the app or removed in response to a verified request.

Sign-in state is stored for your browser session, and voice preferences are stored in browser local storage. Google sign-in may also use Google’s own cookies. WeTalk Money does not add advertising trackers or a separate analytics service to the current app.

We use HTTPS and account-scoped access checks. The app does not intentionally write financial excerpts, request bodies or authentication tokens to application error logs. Infrastructure providers maintain their own operational records. No method of storage or transmission can be guaranteed completely secure.

Deleting financial data removes the active financial context, retained snapshots and saved sessions for your account. It does not delete your Google account, the Firebase sign-in account record, downloaded files, rate-limit counters or information held by a speech provider. Provider logs and any infrastructure backups follow their applicable retention arrangements; contact us about account removal or retained copies.

Your choices and requests

• Use Sources & imports to review conversation-storage and cloud-processing permissions, manage imported records and download your context.
• Use Session analysis log to download retained imported text and selected or all saved conversations and permitted CAF action observations. This export contains private information; you choose whether and with whom to share it.
• Use Delete my financial data to request deletion of stored financial context, snapshots and sessions. The app asks you to confirm before deleting.
• Use Settings to choose voices and control spoken replies, dictation and optional voice processing. You can also revoke microphone permission in your browser.

Contact us to request access, correction, account removal or help with deletion and retention. We may need to verify that a request belongs to the account holder. Any additional privacy rights depend on the laws that apply to you.

Contact and updates

For privacy questions or requests, contact clinton@livewebflyer.com. Please describe your request without emailing complete financial documents, passwords or authentication codes.

We may update this notice as the service changes. The effective and last-updated dates above identify the current version.