Privacy Policy

Interview AI Copilot: Whisper (iOS app)

Effective date: 29 October 2026
Last updated: 29 October 2026

This policy applies to version 4.0 and later of the Interview AI Copilot: Whisper app ("the app"). It explains what data the app processes, how the data is collected, what it is used for, who receives it, where it is processed, how long it is kept, and how you can withdraw your permission and delete your data.

1. Who we are

The app is published on the App Store by Dmitrii Gusev, an individual developer ("we", "us"). We are responsible for the processing described in this policy.

• Contact: aiinterviewcopilot@gmail.com
• Website: https://www.aiinterviewcopilot.com

The app has no user accounts. We do not run servers that store your conversations, transcripts, interview records, Prep data, résumé or images. Our only server component is a Google Cloud function that gives the app access keys for the AI services. It does not receive your content (section 3.13).

2. Summary

• The app helps you before and during job interviews. The prompter listens to the interview and whispers hints into your earphone. Copilot (formerly "Chat") answers questions by text and voice. Prep keeps what you tell the app about the interview and about yourself.
• To do this, the app sends data from your device directly to third-party AI services: OpenAI, Anthropic, Z.ai (Zhipu AI), DeepSeek, Deepgram, ElevenLabs, xAI and Tencent Cloud. This includes:
  • your voice and, during a prompter session, everything the microphone picks up, including what the interviewer and other participants say;
  • your messages, transcripts and your Prep data, for example your résumé with the name and contact details in it, your stories, your salary expectations and the files you mark for use in answers;
  • photos, screenshots and pasted images;
  • the text of hints and answers that are read aloud.
• New installations of the app ask for your permission on a "Your data and AI" screen during onboarding or before the first AI action. Until you tap Allow, the app does not send your voice, text, images or other content to these services. If you updated from version 3.x, the app did not show you this screen, including for the new kinds of data version 4.0 sends (section 10.2). You can review what is sent and turn data sharing off at any time in Settings → General.
• Deepgram, ElevenLabs and OpenAI may keep the data they receive and use it to improve and train their models. DeepSeek stores data in the People's Republic of China, and its API terms contain no commitments on retention, model training or security of API data (section 9).
• Your Copilot session history, Prep data, interview records and settings are stored on your device and, if you use device backups, in your backup (iCloud or a computer). We do not store them on our servers.
• The app also uses Google Firebase (analytics, app integrity checks, remote configuration and a cloud function) and Sentry (crash and error reporting). Error reports can contain short fragments of recognized speech, of what you type and of AI answers, and the country of your App Store account (section 3.11).
• We do not sell your personal data. The app shows no ads, and we do not use your data for advertising.

3. Data we process and how we collect it

We collect data in three ways:

• From you, when you use the app: what the microphone picks up while you record or run a prompter session, what you type or enter in Prep, the files, images and text you add or paste, and what you copy to the clipboard while automatic pasting is on.
• Automatically, from your device: technical, diagnostic and usage data (sections 3.11–3.15).
• From the services we use: transcripts, hints, answers, image descriptions and synthesized speech that the AI services return to the app.

3.1 Microphone audio

• Prompter. From the pre-session check (sound calibration) until you tap End, the app captures everything the device microphone picks up — your voice and the voices of other participants, such as the interviewer — and streams it in real time to Deepgram for transcription. Deepgram is used for the prompter in every region. The session keeps running when you switch apps or lock the screen, and it does not stop on its own after silence. It ends when you tap End, when you withdraw your permission (section 10.4), or after 4 hours at most. iOS shows its microphone indicator while the microphone is in use. The app does not save the audio. The transcript becomes part of the interview record (section 8).
• Copilot, Live mode. After you start recording, the app streams microphone audio to the speech-recognition service selected in Settings → Copilot → Transcription Provider: Deepgram (the default), ElevenLabs Scribe, or OpenAI Realtime (not available in the China storefront). The app does not save Live audio. If no speech is detected for 15 minutes while the app is in the background, recording stops automatically.
• Copilot, Manual mode. The app records audio to a file on your device. When you send a recording, the file is uploaded to ElevenLabs for transcription. The recording files stay on your device until you end the Manual recording with the Stop button or withdraw your permission during it, which deletes them. If the app is closed during a recording, the files can remain until the next time you end a Manual recording with Stop, or until you delete the app.
• Deepgram, ElevenLabs and OpenAI may keep the audio they receive — including other participants' voices — and use it to improve their services and to train and test their models (section 9). OpenAI receives audio only from Copilot Live recordings when OpenAI Realtime is selected.
• The app uses the microphone only after you grant the iOS microphone permission.
• The prompter records other people's speech. Please read "Other people's speech" in section 8.

3.2 Recognized speech (transcripts)

• The speech-recognition service returns text. In Copilot, the app shows it and stores it in your session history on the device. In the prompter, it is stored in the interview record (section 8).
• Copilot, Live mode. The app regularly sends the latest part of the transcript (about the last 300 characters) to a language model to detect interview questions. Outside the China storefront this is OpenAI; if it is unavailable, DeepSeek, then Z.ai. In the China storefront it is DeepSeek, then Z.ai. When a question is detected, the app sends it automatically after a short countdown (3 seconds by default; you can change it or turn it off in Settings → Copilot → Question Display Delay) as a Copilot message, together with the context described in section 3.3.
• Prompter. Each request for a hint includes the recent part of the transcript — usually the last 90 seconds — including words that are still being recognized (section 3.5).

3.3 Copilot messages and the context sent with them

Each Copilot request to a language model can contain:

• the message you type or send, or the transcript or question sent in Live or Manual mode;
• recent messages of the current session as context;
• your Prep data (section 3.4);
• settings that shape the answer, such as the answer language, and today's date.

You choose the Copilot model in Settings → Copilot → Language Model: models from OpenAI, Anthropic, Z.ai or DeepSeek. In the China storefront, only Z.ai and DeepSeek models are available.

3.4 Prep data

What you can enter on the Prep tab:

• Interview: company, role, interview type, seniority, tech stack, the job description text and its link, the interview language, the programming language for tasks, and questions you want to ask the company.
• About me: your résumé; your stories (situation, task, actions, result, numbers, lessons, how people reacted); your facts — salary expectations (amount, currency, period, gross or net, variable part), notice period, start date, relocation, work format, work authorization, other offers and your reason for changing jobs; your motivation (growth goals, strengths and growth areas, what motivates you); how the app should refer to you in languages with grammatical gender; additional context; and files.
• Answer rules: built-in and your own rules for how answers should sound.

Résumé. When you choose a PDF résumé, the app extracts its text on your device, using text recognition for scanned pages if needed. It stores the text in the app's settings and saves a copy of the PDF file in the app's documents folder. When you delete the résumé in Prep → About me, the app also deletes the saved copies of the PDF file.

Files. You can add PDF, TXT or Markdown files. The app extracts their text on your device and keeps only the file name and the text (up to 40,000 characters per file, up to 10 files), not the file itself.

What is sent to AI services. Your Prep data is sent to the language model as part of Copilot requests (section 3.3) and prompter requests, including when a prompter session starts (section 3.5):

• the answer rules that are turned on. Some built-in rules are on by default, so they are sent with every Copilot and prompter request, even if you have not filled in anything in Prep;
• the job: company, role, interview type, seniority, tech stack, the job description text, the interview language and the programming language for tasks, when you have filled them in;
• the résumé text only while "Use in answers" is on for it;
• file texts only for files with "Use in answers" turned on;
• your stories, your facts (including your salary expectations), your motivation, how the app should refer to you in languages with grammatical gender, additional context and your questions for the company, when you have filled them in.

The job link itself is not sent to AI services. Prep data is sent as you entered it: the app does not remove names or contact details from it. For example, the name, email address and phone number in your résumé are sent as part of its text.

Build from résumé. When you tap Build from résumé, the text of your résumé (up to 40,000 characters) is sent to the default language model for your region — OpenAI outside the China storefront, DeepSeek in it — or, if that service refuses the request, to another one (section 5.2). The model returns draft stories. They are kept only if you save them.

Loading a job description by link. When you tap Load, the app requests the page directly from the website at that link. The website receives your IP address and standard request information, as with any web browser. The link is not sent to us or to AI services. The app extracts the text of the page on your device.

3.5 Prompter requests

During a prompter session, each request for a hint sent to the language model can contain:

• the recent part of the transcript (section 3.2);
• your Prep data (section 3.4);
• the session's question journal: the questions understood so far, short summaries of the hints already given, which of your stories have been used, and the titles of attachments;
• attachments (section 3.6): those pinned for the whole session and those of the current question;
• your command (for example, Help, Shorter, Example, Translate) and settings such as the hint language, and today's date.

At the start of a session and while it runs. When you tap Start, the app sends a preparation request with your Prep data, without the transcript. It repeats this request during the session after a period without requests, together with the attachments you pinned for the whole session. This prepares the model and keeps the connection ready so that hints arrive faster. This means your Prep data reaches the language model even if you never ask for a hint.

Which model. Outside the China storefront, OpenAI. In the China storefront, a Z.ai or DeepSeek model. If the main service is unavailable or refuses the request, the request goes to another one (section 5.2).

The text of hints is read aloud as described in section 3.8.

3.6 Photos, screenshots, images and other attachments

• Copilot. When you take a photo with the camera, choose an image in the photo picker or paste an image, the app compresses it and sends it right away, together with the session context (section 3.3), to an image-capable model: OpenAI or Anthropic, if you selected one of their models; Z.ai, if you selected a Z.ai or DeepSeek model, and as a fallback for any image. A small thumbnail is stored in your session history on the device.
• Prompter. You can paste text or code from the clipboard, take a photo of a screen, choose a photo, or add the text of a PDF (extracted on your device; a scanned PDF is sent as an image). Attachments are sent with the next request for a hint (section 3.5). An image is sent once; after the model has read it, the app sends the model's description of the image instead. Images are read by OpenAI outside the China storefront and by Z.ai in it; when a DeepSeek model answers instead of the main model, images go to Z.ai. The texts of attachments and copies of the images (reduced to at most 2048 pixels on the long side) are saved in the interview record (section 8).
• The app uses the camera only with your iOS permission. From your photo library, the app receives only the images you choose in the system picker.

3.7 Clipboard

• Copilot — Auto-paste Images is on by default. While the app is open, when you copy an image — including through Universal Clipboard from another Apple device, such as your Mac — the app adds it to the current Copilot session, or starts a new one, and sends it to an image-capable model as described in section 3.6, without an extra tap. iOS may ask you to allow pasting. An image you copy while the app is in the background is not read or sent.
• Copilot — Auto-paste Text is off by default. If you turn it on, text you copy while a Copilot session is open on the screen is sent to the language model as a question in that session. Text you copy elsewhere in the app is not read. Text you copy in another app while the app is in the background is read and sent when you return only if a Copilot session is open at that moment, and only as text.
• You can turn either option on or off in Settings → Copilot.
• Prompter — Auto-paste from Mac is off by default. While it is off, the prompter only checks whether the clipboard has changed, without reading its contents, and reads it only when you tap Paste. If you turn it on, text and images you copy during a session are added and sent automatically, except short copies, repeats and strings that look like passwords or keys, which the app offers to send with a tap.
• Until you allow data sharing with AI services (section 10), nothing from the clipboard is sent to them.

3.8 Text for voice playback

• Copilot. When you tap Play on an answer or turn on automatic playback (off by default), the text of the answer and its language are sent to a text-to-speech service: xAI, or ElevenLabs if xAI is unavailable. In the China storefront, Tencent Cloud is used. The synthesized audio is saved on your device together with the answer text, so that you can replay the answer (section 7.1).
• Prompter. The text of each hint is sent sentence by sentence to a text-to-speech service: xAI first, and ElevenLabs if xAI is unavailable or slow to start playing. In the China storefront, Tencent Cloud is used. Hints can contain details from your Prep data, for example numbers from your stories or your salary expectations. The synthesized audio is kept in a temporary folder during the session, so that hints can be replayed, and the folder is deleted when the session ends.

3.9 Interview records

After a prompter session, the app keeps a text record of it on your device. See section 8.

3.10 What the app does without AI services

The onboarding rehearsal, lessons and cheat sheet work without the microphone, the network or AI services; the rehearsal plays recordings built into the app. Editing Prep data, reading your Copilot session history and interview records, changing settings and purchasing work without sending anything to AI services.

3.11 Crash and error reports (Sentry)

Sentry collects automatically:

• crash reports, errors and app hangs, with technical details such as the stack trace, device model, operating system and app version;
• session health information and performance data for a 10% sample of app operations, and technical summaries of prompter sessions (for example, speech-recognition delays), without text;
• a random identifier created for your installation of the app, which is attached to every report, and your subscription status (free or Pro). For errors during purchases, the product identifier is included; for errors when a purchase is checked, the transaction identifier and the subscription expiration date are included.

Each error report includes up to 50 of the most recent log entries ("breadcrumbs"). These entries can contain short fragments of text that the app processed — for example, recognized speech (which can include other participants' words), detected interview questions, the text in Copilot's message field and fragments of AI answers — and the country of your App Store account.

Error reports can also include error messages returned by the AI services (up to about 1,000 characters) and technical descriptions of errors when the app saves data on the device. Such messages can occasionally contain parts of the request or of the data being saved.

For 10% of sessions in which an error occurs, Sentry also records a session replay: a screen recording of up to 30 seconds before the error, continuing until the session ends. The Sentry SDK masks all text, images and user input on the device before a replay is sent. The app does not send screenshots or view hierarchies to Sentry. A limited number of unsent reports can be stored on your device until they can be sent.

3.12 Usage analytics (Google Analytics for Firebase)

Google Analytics for Firebase collects automatically:

• an app-instance identifier and Apple's identifier for vendor (IDFV), which are attached to every event;
• device model, operating system version, app version and language;
• approximate location (for example, country or city) that Google derives from your IP address;
• usage events, such as first open, sessions, screens viewed and engagement time;
• events we define, for example: onboarding steps; opening the subscription screen and where it was opened from; completed purchases with the product identifier; prompter session statistics such as duration, number of hints, which controls and commands were used, response times and error types; review prompts and whether your response was positive or negative; and interactions with in-app promotions.

Google Analytics may also automatically record in-app purchase details, such as the product identifier, name and price. Analytics events do not contain your conversations, transcripts, Prep data, résumé, interview records or images.

The app does not ask for tracking permission (App Tracking Transparency). Without that permission, iOS does not give apps the advertising identifier (IDFA), so the Google Analytics SDK cannot collect it. The app also includes Google's on-device conversion measurement library for Google Analytics, but the app does not call it.

3.13 App integrity, access keys and configuration (Google Firebase)

• App Check uses Apple's App Attest or DeviceCheck to verify that requests come from the genuine app. It processes attestation data and a short-lived token.
• Our cloud function runs on Google Cloud in the United States. It receives an empty request together with the App Check token and returns access keys for the AI services. It receives no content from you. Google Cloud records technical request logs, such as the IP address.
• Remote Config and Firebase Installations process a Firebase installation identifier, the app version, the operating system version and your device's language, country code and time zone. We use Remote Config to show in-app messages, such as update notices and promotions.

3.14 Purchases and data from Apple

• Subscriptions and purchases are processed by Apple through the App Store. We do not receive your name, Apple ID or payment details. The app checks your purchase status on the device with Apple's StoreKit to unlock Pro features, and reads the country of your App Store account on the device to choose the services available in your region. The app writes this country to its log, so it can reach Sentry with an error report (section 3.11).
• If you have chosen in iOS settings to share analytics with app developers, Apple may provide us with crash reports and aggregated usage statistics under Apple's terms.
• To decide whether to show a promotion for the Mock Interview AI app, the app checks on your device whether that app is installed.

For purchase-related data in analytics and diagnostics, see sections 3.11 and 3.12.

3.15 Technical data

Every online service that the app contacts receives your device's IP address and basic technical information about the request. The app has no fields for your name, email address or phone number and does not add an account or device identifier to requests to the AI services. Your content, however, is sent as it is: your résumé usually contains your name, email address and phone number and is sent while "Use in answers" is on (section 3.4), and other content — Prep data, a photo or the conversation — can contain such details too.

3.16 Emails to us

If you email us, we receive your email address and whatever you include in the message. We use it only to reply to you.

4. How we use data

We use data to:

• provide the app's features: prompter hints, speech transcription, interview-question detection, Copilot answers, image and attachment analysis, voice playback, story drafts from your résumé, loading job descriptions, and interview records;
• keep features working when a service fails, by switching to another service (section 5.2);
• protect access to the AI services and prevent abuse (App Check and access keys);
• find and fix crashes and errors, and improve stability and performance (Sentry);
• understand how the app is used in aggregate and improve it (Google Analytics);
• show in-app notices, such as update notices and promotions (Remote Config);
• unlock purchases and Pro features;
• reply to your messages and requests;
• comply with legal obligations.

We do not sell your personal data, do not use it for advertising, and do not train AI models on it. How Deepgram, ElevenLabs and OpenAI may use the data they receive is described in section 9. The app makes no decisions about you that have legal or similarly significant effects.

5. Who receives data and why

5.1 AI services

5.2 Automatic switch to another service

If the service that should handle a request is unavailable or refuses it, the app can send the same request, with the same data, to another service from the table above, so that you still get a result:

• Copilot answers and Build from résumé: outside the China storefront, DeepSeek, then Z.ai; in the China storefront, the other available model (DeepSeek or Z.ai);
• interview-question detection in Copilot: as described in section 3.2;
• prompter hints: outside the China storefront, DeepSeek, then Z.ai; in the China storefront, the other Z.ai or DeepSeek models, also when Z.ai reports too many requests;
• images: Z.ai;
• Copilot voice playback outside the China storefront: ElevenLabs;
• prompter voice playback outside the China storefront: ElevenLabs, also when xAI is slow to start playing.

When another model answers in Copilot instead of the one you selected, the app shows a notice. The prompter switches without a notice. This means a request can reach a service from the list that you did not select.

5.3 Other recipients

• Google (Firebase): Google Analytics for Firebase, App Check, Remote Config, Firebase Installations and Cloud Functions, as described in sections 3.12 and 3.13.
• Sentry: crash and error reporting, as described in section 3.11.
• Apple: purchases and subscriptions through the App Store, App Attest and DeviceCheck. If you use iCloud Backup, the data the app stores on your device is included in your backup under Apple's terms, except the audio of Copilot answers read aloud (section 7.2).
• Websites you load job descriptions from: only when you tap Load (section 3.4).
• Authorities: we may disclose data if the law requires it.
• A new owner: if the app is transferred to another person or company, this policy continues to apply to the data processed before the transfer.

6. Where data is processed

This means your data can be transferred to countries whose data-protection laws differ from those where you live, including the United States, Singapore and China. The terms of several of these providers include safeguards for international transfers, such as standard contractual clauses or the EU–U.S. Data Privacy Framework.

7. How long data is kept

7.1 On your device

• Copilot session history (messages, transcripts, AI answers and image thumbnails): until you delete a session, clear the session history, or delete the app. There is no automatic deletion.
• Prep data: until you change or delete it, or delete the app. Clearing the session history does not delete it.
• Résumé: the extracted text and the saved copies of the PDF file until you delete the résumé in Prep → About me or delete the app.
• Interview records: according to the setting you choose (section 8).
• Manual-mode recordings: as described in section 3.1.
• Audio of Copilot answers read aloud, with the answer text: kept so that you can replay answers. It is not included in device backups, is limited to about 200 MB (the oldest audio is removed first), and audio of deleted sessions is removed the next time the app starts.
• Audio of prompter hints: deleted when the session ends.
• Your data-sharing choice, settings and prompter settings: until you change them or delete the app.
• Unsent crash reports: until they are sent.
• Access keys for the AI services are kept in the iOS Keychain. They are not your personal data.

7.2 Device backups

If you back up your device (iCloud Backup or a computer), the data the app stores on the device — including Copilot session history, Prep data, the résumé text and PDF copy, interview records with their images, Manual-mode recording files that are still on the device, your data-sharing choice and settings — is included in the backup and kept according to your backup settings. The audio of Copilot answers read aloud is excluded from backups. Restoring a backup restores this data.

7.3 At service providers

These are the periods the providers state in their terms. We do not control them.

8. Interview records

When you use the prompter, the app saves a text record of the session: the recognized conversation — your words and the words of the other participants — the questions, the hints, the attachments (their texts and copies of images) and translations you requested, together with basic Prep details (company, role, interview type and languages). The audio is not saved.

Where it is stored. The record is stored on your device, in the app's database, with the data protection iOS applies to app data by default. Copies of images in the record are saved with iOS data protection that keeps them encrypted until you first unlock the device after it restarts. The record is included in your device backup (iCloud or a computer) if backups are on. The record is not sent to our servers, and the app does not send the record itself to anyone.

What AI services receive during a session is described in sections 3.1 and 3.5: the microphone audio, continuously; your Prep data when the session starts and, together with pinned attachments, periodically while it runs; and the transcript, question journal and attachments with each request for a hint.

How long it is kept. You choose how long records are kept in Settings → Prompter → Keep interview records: Always (the default), 30 days, or Don't keep.

• 30 days are counted from the start of the session. The app checks the period when it starts and when you change the setting. On the first launch after you restore the app from a backup, the 30-day limit is applied from the next launch.
• Don't keep: the record of the session is still made and normally shown after you tap End. It is deleted when you close its screen or, if the app closes before that or deleting fails, the next time the app starts. If its screen doesn't open after the session, the record is deleted at once.
• If you change the setting in a way that would delete existing records, the app asks you first.

How to delete records. You can delete one record on the Past sessions list on the Prompter tab (swipe it and confirm) or on its screen, and all records with Settings → Prompter → Delete all records. Clearing the session history does not delete interview records. Deleting the app deletes them. A backup made before the deletion can still contain a record, and deleted records can remain for some time in the internal files of the app's storage on the device.

Other people's speech. A prompter session captures and transcribes everything the microphone picks up, including what the interviewer and other participants say. Their speech is sent to Deepgram for transcription, may be used by Deepgram to improve its models (section 9), and is kept in your interview record. In some countries and states, recording or transcribing a conversation requires the consent of everyone taking part. Before you use the prompter, check the rules that apply to you and inform the other participants where required.

9. How third parties protect your data

We share personal data only with the recipients named in section 5 and only for the purposes in section 4. We use each service as a business customer under its published terms for business or API customers, and we reviewed those terms when preparing this policy. For each recipient, below is what protects your data and what that protection rests on.

• OpenAI. Basis: OpenAI's data processing addendum, which applies when its services are used, with standard contractual clauses for transfers out of the European Economic Area. Data sharing with OpenAI is turned on in our OpenAI account, so OpenAI may retain requests and may use them to train its models. Retention is described in section 7.3.
• Anthropic. Basis: Anthropic's data processing addendum, which is part of its commercial terms. Under those terms, Anthropic may not train models on customer content from its commercial services. Retention is described in section 7.3.
• Z.ai. Basis: Z.ai's data processing addendum for API services, under which Z.ai acts as a data processor on the customer's instructions. Z.ai will not use API content to develop or improve its services unless the customer explicitly agrees. Z.ai states that API content is processed in real time and not saved on its servers.
• xAI. Basis: xAI's published security terms for its API: requests and responses are stored encrypted at rest for 30 days for abuse auditing, and xAI does not train on API inputs or outputs without the customer's explicit permission.
• Tencent Cloud. Basis: Tencent Cloud's data processing and security agreement, which applies together with its terms of service. Under it, Tencent Cloud processes personal data only to perform the service, following the customer's instructions, in the region the customer selected (Singapore).
• Deepgram. Basis: Deepgram's terms and privacy policy, under which Deepgram processes customer data on behalf of its customers and applies technical and organizational security measures. The same terms allow Deepgram to use the content sent to its API — in the app, the microphone audio of prompter sessions, including other participants' voices, and of Copilot Live recordings when Deepgram is selected — to improve its services, including training and testing its models, unless a request opts out of its Model Improvement Program. The app does not opt out, so Deepgram may store portions of this audio for model improvement, with no retention period stated. See https://deepgram.com/terms and https://developers.deepgram.com/docs/the-deepgram-model-improvement-partnership-program.
• ElevenLabs. Basis: ElevenLabs' data processing addendum for self-serve accounts, which requires technical and organizational security measures and processing only on the customer's documented instructions, which include ElevenLabs' terms of use. Those terms allow ElevenLabs to use submitted content — in the app, Copilot audio (Live recordings when ElevenLabs Scribe is selected, and Manual recordings), the text of Copilot answers read aloud, and the text of prompter hints, which can contain details from your Prep data — to improve its services and to train its models, unless the account holder opts out. We have not opted out. ElevenLabs keeps a history of requests until it is deleted (section 7.3). See https://elevenlabs.io/terms-of-use and https://elevenlabs.io/privacy-policy.
• DeepSeek. Basis: only DeepSeek's terms of service for its open platform and its privacy policy. We did not find a data processing agreement. DeepSeek's API terms contain no commitments on how long DeepSeek keeps the data the app sends, whether it uses that data to train models, or how it secures that data, and they make the app developer responsible for end users' personal information. DeepSeek's privacy policy describes security measures for DeepSeek's own services but states that it does not cover end users of apps built on DeepSeek's platform. DeepSeek stores data in the People's Republic of China. See https://cdn.deepseek.com/policies/en-US/deepseek-open-platform-terms-of-service.html and https://cdn.deepseek.com/policies/en-US/deepseek-privacy-policy.html.
• Google (Firebase). Basis: Google's terms for Firebase, under which Google generally acts as a data processor for its customers. Google Analytics data is processed under the Google Analytics terms.
• Sentry. Basis: Sentry's data processing addendum, under which Sentry acts as a data processor, and its AI and machine-learning policy: Sentry does not use customer data to train generative AI models without permission; it may use service data to train models for features such as issue grouping, and states that it removes personal information before data enters a training set.
• Apple. Basis: Apple's privacy policy, for App Store purchases, App Attest and DeviceCheck requests and iCloud backups.

Our confirmation. On the basis listed for each recipient above, we confirm that each recipient named in section 5 protects your data at a level that is the same as or equal to the protection described in this policy. The protection this policy describes for each recipient is the one listed above, including these limits, which we disclose so that you can decide before you tap Allow:

• OpenAI may retain requests, with their full text, in the logs of our OpenAI project and may use them to train its models, because data sharing with OpenAI is turned on in our account. This covers everything OpenAI receives from the app, including the résumé text with the contact details in it and images.
• Deepgram's Model Improvement Program covers the audio of prompter sessions, including other participants' voices, and of Copilot recordings sent to Deepgram.
• ElevenLabs may use submitted content, including the text of prompter hints, to improve its services and train its models, and it keeps request history until it is deleted.
• For DeepSeek, this confirmation rests only on its published terms and privacy policy. There is no contractual commitment on retention, model training or security of the data the app sends, and the data is stored in China.

Before you allow data sharing, the permission screen in the app names the AI services that receive your data and states that Deepgram, ElevenLabs and OpenAI may keep the data they receive and use it to improve and train their models.

10. Your permission and how to withdraw it

10.1 New installations

• When the app asks. During onboarding, after the rehearsal, the app shows a step titled "Your data and AI". If you skip onboarding, leave it before this step, or tap Not Now, the same screen appears before your first AI action — for example, starting a recording, sending a message, retrying an answer, adding or pasting an image or playing an answer in Copilot, starting a prompter session, or tapping Build from résumé or Retry in its story drafts.
• What the screen says. It explains that, to listen to your interview and help you answer, the app sends data to third-party AI services when you use the prompter, Copilot and Prep. It lists what is sent and which services receive it:
  • your voice, for transcription — the prompter listens to the whole conversation, including the interviewer; Copilot, while recording: Deepgram, ElevenLabs, OpenAI;
  • text, for hints, answers and story drafts — transcripts, your messages, session questions, pasted text and your Prep data: answer rules, job, résumé, stories, facts (including salary), motivation, additional context, your questions for the company, grammatical gender and files marked "Use in answers": OpenAI, Anthropic, DeepSeek, Z.ai;
  • images, for task analysis — photos, screenshots and pasted images: OpenAI, Anthropic, Z.ai;
  • the text of hints and answers, for voice playback: ElevenLabs, xAI, Tencent Cloud.

  It also states that Deepgram, ElevenLabs and OpenAI may keep the data they receive and use it to improve and train their models, that nothing is sent to these services until you tap Allow, and that Copilot sends copied images automatically, which you can turn off in Settings → Copilot; when Auto-paste Text is on, or when you turned it on earlier in onboarding, the note also says that Copilot sends text you copy in an open session. A separate note on the screen says that error reports go to our diagnostics service Sentry whether or not you tap Allow, and that they may contain fragments of your text, transcripts and AI answers, and your App Store country (section 3.11). The promise that nothing is sent until you tap Allow applies to the AI services, not to these diagnostic reports. The screen links to this policy.

• Your choice. The screen has two buttons of equal weight, Allow and Not Now, with nothing preselected. Until you tap Allow, the app does not send your voice, text, images or other content to these AI services. By tapping Allow, you agree to the data sharing described on the screen and in this policy, including the use of data by Deepgram, ElevenLabs and OpenAI described in section 9.
• If you tap Not Now during onboarding, onboarding continues. If you tap Not Now or swipe the screen away before an AI action, that action is cancelled, the app tells you where to turn data sharing on later, and the rest of the app keeps working. AI features are unavailable without your permission. The screen appears again the next time you use an AI feature.

10.2 If you updated from version 3.x

Version 3.x of the app already sent data to these AI services: your voice while recording, transcripts, Copilot messages with recent session history, your prompts and résumé, images, and the text of answers read aloud. Version 4.0 sends new kinds of data:

• during a prompter session, everything the microphone picks up, including what the interviewer and other participants say;
• your Prep data, such as your stories, your salary expectations and the files you mark for use in answers;
• prompter attachments and the text of prompter hints read aloud.

If you updated from version 3.x, the app treats data sharing as allowed and, in most cases, does not show you the permission screen. This means the app did not ask your permission for these new kinds of data either.

After the update, the What's New screen includes the item "What data goes to AI services", which opens the description of what is sent and to whom. You can review it at any time in Settings → General → What is sent and to whom. In the same place, you can turn data sharing off with the Send data to AI services switch (section 10.4).

10.3 Your choice is stored on your device

Your choice is stored on your device, not on our servers. If you delete and reinstall the app, it asks again. If you restore the app from a backup, your choice is restored with it.

If a future version of the app changes the types of data, the recipients or the purposes shown on the permission screen, earlier permission stops counting — including permission carried over from version 3.x. The app then shows the permission screen to everyone before the next AI action.

10.4 How to withdraw your permission

You can withdraw your permission at any time in Settings → General by turning off Send data to AI services. Withdrawing takes effect immediately:

• the app stops any recording, answer generation, voice playback and automatic pasting in progress;
• a prompter session in progress ends as if you had tapped End, and a pre-session check closes;
• a Manual recording in progress ends as with the Stop button (section 3.1).

After that, the app does not send data to AI services until you allow it again, on the permission screen or with the same switch. Withdrawing your permission does not delete data that is already stored on your device or was already sent (sections 7 and 11).

10.5 Other choices

• Auto-paste Images in Copilot is on by default. You can turn it off in Settings → Copilot → Auto-paste Images.
• Declining or withdrawing your permission does not cancel a subscription, and purchases do not require this permission. AI features, including prompter sessions, need it. You can manage subscriptions in your Apple ID settings.
• You can change microphone and camera access in iOS Settings → Privacy & Security.
• Usage analytics and crash reporting do not depend on this permission. The app currently has no in-app setting to turn them off. You can contact us with questions or requests about this data.

11. Deleting your data

• In the app:
  • to delete one session, swipe it left in the session list on the Copilot tab and tap Delete;
  • to delete all sessions, use Settings → Copilot → Clear Session History (your settings, Prep data and interview records are kept);
  • interview records: see section 8;
  • Prep data: edit or clear the fields, and delete stories and files in Prep → About me;
  • to remove your résumé, delete it in Prep → About me. This removes the extracted text and the saved copies of the PDF file (section 7.1).
• Deleting the app removes all data the app stored on your device. Copies in your device backups remain according to your backup settings.
• At service providers: the app does not send an account identifier to the AI services, so we usually cannot find the records that belong to you. These records are deleted according to section 7.3. You can contact us with details, such as the approximate date and time of use, and we will pass your request on to the provider where the provider's procedures allow it. You can also contact the providers directly (section 16).
• Analytics and crash reports are associated only with random installation identifiers, so we usually cannot match them to you. Contact us and we will help where we can.

12. Your rights

Depending on where you live — for example, in the European Economic Area, the United Kingdom or California — you may have the right to access, correct or delete your personal data, to restrict or object to its processing, to data portability, and to withdraw your consent. You may also have the right to complain to a data-protection authority.

Because the app has no accounts and your Copilot session history, Prep data and interview records are stored on your device, not on our servers, you can exercise many of these rights directly in the app (sections 10 and 11). For other requests, email us at aiinterviewcopilot@gmail.com. We may ask for details that help us locate the data, because we cannot otherwise identify you.

If your voice was captured by someone else's prompter session, the transcript is stored on that user's device, not on our servers, and the audio was processed by Deepgram as described in sections 7.3 and 9. You can contact us, and we will help where we can.

Where the EU or UK GDPR applies, we rely on:

• your consent to send your content to AI services — you can withdraw it at any time (section 10);
• performance of a contract to provide the features you use and your purchases;
• our legitimate interests in keeping the app secure and stable and in understanding its use in aggregate (App Check, crash reports and analytics);
• legal obligations, where the law requires us to process or disclose data.

We do not sell your personal information and do not share it for cross-context behavioral advertising.

13. Children

The app is designed for adults preparing for and taking job interviews and is not directed at children. We do not knowingly process personal data of children under 13, or under the higher age of digital consent that applies where they live. If you believe a child has used the app and sent personal data through it, contact us and we will help with the steps in section 11.

14. Security

The app sends data to all services over encrypted connections (HTTPS or secure WebSockets). Access keys for the AI services are delivered through a cloud function protected by Firebase App Check and are stored in the iOS Keychain. Data that the app stores on your device is protected by iOS security features; copies of images in interview records are kept encrypted until you first unlock the device after it restarts. No method of transmission or storage is completely secure.

15. Changes to this policy

When we change this policy, we will update this page and the "Last updated" date. If a future version of the app starts sending new types of data to AI services, adds a new AI service or changes the purposes shown on the permission screen, we will update this policy, and the app will ask everyone for permission again before the next AI action, as described in section 10.3.

16. Contact and provider policies

• Email: aiinterviewcopilot@gmail.com
• Developer: Dmitrii Gusev

Privacy information from the services we use:

• OpenAI: https://openai.com/policies/privacy-policy/ and https://developers.openai.com/api/docs/guides/your-data
• Anthropic: https://www.anthropic.com/legal/privacy and https://privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-organization-s-data
• Z.ai: https://docs.z.ai/legal-agreement/privacy-policy
• DeepSeek: https://cdn.deepseek.com/policies/en-US/deepseek-privacy-policy.html
• Deepgram: https://deepgram.com/privacy
• ElevenLabs: https://elevenlabs.io/privacy-policy
• xAI: https://x.ai/legal/privacy-policy
• Tencent Cloud: https://www.tencentcloud.com/document/product/301/17345
• Google Firebase: https://firebase.google.com/support/privacy
• Sentry: https://sentry.io/privacy/
• Apple: https://www.apple.com/legal/privacy/