Privacy Policy

Data we process

When you sign in, we receive an account identifier from Apple or Google and, when the provider shares it, your name or email address. We also process session tokens to maintain authenticated access.

To provide dictation, we temporarily process the audio you stream, resulting transcripts, text sent for cleanup, your custom vocabulary, your replacements and writing styles, your preferences, and a short summary of the technical topics you dictate about (described below). If you turn on sync, we also process your dictation history and the contents of your notes.

We may also receive technical data such as app version, operating system, session and device identifiers, an approximate location derived from your IP address, response times, usage, product events, and error or security logs.

How we use data

We use this data to authenticate you, turn speech into text, improve punctuation and formatting, personalize recognition with your vocabulary and topic summary, maintain access, enforce plan limits, troubleshoot failures, prevent abuse, and secure the service.

We may also use aggregated or de-identified metrics to understand performance and improve Parlate. We do not sell your personal data or use dictation content for targeted advertising.

Providers and transfers

Parlate reaches your device through servers operated by Reserved Technologies, but those servers act as intermediaries: speech recognition and text cleanup are performed by external providers, including third-party artificial intelligence services. Each one is identified below, along with what it receives.

• xAI (Grok) and Soniox — speech-recognition providers that transcribe your dictated audio. xAI (Grok) is the primary provider and Soniox takes over when xAI is unavailable; in some cases both transcribe the same dictation so their results can be combined for better accuracy. Depending on service configuration and availability, they receive the audio and, when used, your custom vocabulary or your topic summary, solely to improve recognition.
• AssemblyAI — failover speech-recognition provider. Receives audio only when the failover mechanism is engaged.
• Groq, Cerebras and Google (Gemini) — artificial intelligence services that receive the already-transcribed TEXT (never the audio) to correct punctuation, formatting, style, and translation. Cerebras and Groq are the primary services; Google (Gemini) acts as a fallback. These services may also process that text transiently to generate your topic summary.
• DigitalOcean and Neon — host our servers and our database. On behalf of Reserved Technologies, they store your account data, subscription state, vocabulary and personalizations, your topic summary and, if you turn on sync, your history and notes. Audio passes through those servers without being kept in a permanent archive.
• Apple and Google — process sign-in and return an account identifier and, when the provider shares it, your name or email address.
• RevenueCat — manages your subscription state and receives a user identifier and purchase history.
• Stripe — processes payment when you subscribe from the desktop app or the web. Receives the payment details you enter on its page and your email; Parlate does not receive or store your full card number.
• PostHog — analytics. In the app it receives usage events, user and device identifiers, and an approximate location derived from your IP address; it never receives your audio or the text you dictate. On parlate.app it receives pageviews, download clicks, approximate country, and device type, with no analytics cookie and without identifying you.
• AppsFlyer — install and usage attribution. Receives installation identifiers and a pseudonymous account identifier to measure how the app is discovered and used. The iOS SDK uses strict mode, without IDFA; we do not send it audio, transcripts or typed text.
• Sentry — crash and performance diagnostics. Receives error and performance reports, stripped of dictated content before they are sent.

Our commitments regarding these providers

We confirm that every third party with which Parlate shares user data is contractually required to provide the same or equal protection of that data as described in this Policy and as required by app store policies. We limit sharing to what is strictly necessary for each function and use channels encrypted in transit.

These providers may operate infrastructure in other countries and additionally apply their own privacy and retention policies. We do not authorize any of them to use your audio, transcripts, or notes to train their own models or for advertising. We may replace a provider with another performing an equivalent function, and will update this list when we do.

Audio, text, and retention

Parlate processes audio as a real-time stream. Pending audio may be held temporarily to recover an interrupted connection and complete dictation; we do not maintain a permanent server-side audio archive. Speech providers may retain or log data according to their contracts, settings, and policies.

Transcribed text is returned to the device or app where you dictate. Your dictation history and your notes are stored by default ONLY on your device. If you turn on sync — an option that is off out of the box and is enabled separately on each device — that content is copied to our servers so you can see it on your other devices.

When history sync is on, you choose the retention period: 1, 7, 30, or 90 days. When it expires, the record is deleted. Synced notes are kept until you delete them. You can turn sync off or erase all server-side history at any time from the app's settings.

We may retain account data, synchronized vocabulary, subscription state, sessions, and technical logs for as long as needed to operate the service, meet legal obligations, resolve disputes, and prevent fraud. We delete or de-identify information when it is no longer needed, subject to backups and legal requirements.

Topic summary to improve transcription

To better recognize the technical vocabulary you use, Parlate derives from your dictations a short summary of the technical topics you tend to talk about — for example, “software development” or “payments and subscriptions” — and stores it on our servers, linked to your account. It is a short list of topics, not a copy of what you dictate: we do not store the text of your dictations for this feature.

The summary is sent to the speech-recognition providers along with the audio of each dictation, solely to improve accuracy. We do not use it for advertising or share it for any other purpose. This feature is being rolled out gradually to all accounts.

The summary may be updated over time as you dictate, and it is deleted when you delete your account. You can also ask us for a copy or ask us to delete it by emailing soporte@reserved-technologies.com.

The Parlate keyboard and Full Access

Parlate includes a system-wide keyboard. iOS requires you to enable “Allow Full Access” for that keyboard to reach our transcription server; without that permission the keyboard installs but dictation does not work.

The keyboard processes the selected word and nearby cursor context on your device to offer corrections and suggestions. It does not keep a typing history or send that text to our servers or providers. During dictation it may send a list-format hint — its type and next number — without the existing written content. Audio is transmitted only during dictation you initiate; contentless operational events, such as the start or end of a dictation, are also sent.

Local storage and security

The app may store preferences, vocabulary, and session credentials on your device using secure storage mechanisms available from the operating system. Information you insert into other apps is also subject to those apps' and your device's practices.

We use reasonable technical and organizational safeguards, including encryption in transit, access controls, and revocable sessions. No system is completely secure; protect your device, keep software current, and notify us of suspicious activity.

Your rights and controls

You may request access, correction, deletion, or a copy of your data, and may object to or restrict certain processing where applicable law allows, by emailing soporte@reserved-technologies.com. We will verify your identity before completing requests that affect an account.

You can withdraw your consent at any time without leaving the app: turn off history and notes sync in Settings, revoke the microphone permission or the keyboard's Full Access from iOS Settings, and delete your account from Settings → Account. Deleting your account erases your data from our servers, including history, notes, vocabulary, personalizations, and your topic summary. Not using dictation stops all audio and text from being sent to external providers.

You can sign out, withdraw system permissions, and cancel a subscription through the relevant platform. Deleting the app does not automatically cancel a subscription or by itself delete the account.

Children

Parlate is not directed to children who are too young to consent to data processing in their jurisdiction. We do not knowingly collect children's data without legally required authorization.

If you believe a child provided data without authorization, email soporte@reserved-technologies.com so we can review it and take appropriate action.

Changes and contact

We may update this Policy to reflect changes to the product, providers, or legal requirements. We will post the current version and date here and, where appropriate, provide additional notice.

Reserved Technologies is responsible for the Parlate practices described here. For privacy questions or to exercise your rights, contact soporte@reserved-technologies.com.