Skip to content

Trust center

Subprocessors

Last updated: November 2026

These are the companies that process personal data for Aethel so we can run the service. Each one may use the data only to provide its service to us, except Anthropic, which keeps and uses it under its own consumer terms (see its entry).

Current list

  • Google Firebase and Google Cloud

    What they do for us
    Sign-in, our database, website hosting, the server code that runs interviews, and bot protection (reCAPTCHA Enterprise).
    What they receive
    Account details, your profile, interview transcripts, and results. For bot protection, details about your browser and device.
    Location
    United States. Our database is in Google Cloud’s nam5 location: Iowa and Oklahoma, with South Carolina as its witness region. Our server code runs in us-central1 (Iowa). Sign-in runs only in Google’s US data centers. Website pages are served from Google’s worldwide network, and the pages themselves hold no personal data. Server logs are kept for 30 days in Google’s default global log storage, which can be in any Google data center. The bot check (reCAPTCHA Enterprise) is a global Google service, so the details it checks can be handled in any Google data center.Checked October 6, 2026. Sources: Firebase data locations, Firestore locations, Cloud Functions locations, Cloud Logging locations.
  • Google Workspace

    What they do for us
    Our team email inbox. It holds the emails you send to support@aethel.pro, and the copies of support form messages and company requests.
    What they receive
    What you write to us, with your name and email address.
    Location
    No region named yet. Our team inbox is Gmail in Google Workspace. Some Workspace plans can keep email in the United States or Europe with a data region setting. We have not checked the setting for our account yet, so we do not name a country.Checked October 12, 2026. Sources: Google Workspace data regions.
  • OpenAI

    What they do for us
    Writes interview questions, scores interview answers and writes feedback on them, writes what Nelly says in the coach debrief and the voice practice, writes AI feedback on one free practice answer when you ask for it, and suggests role details from a job description a company pastes.
    What they receive
    Profile details you give us or confirm from your resume (such as skills, work history, and a project you describe), your interview transcript, the text of what you say in a coach debrief or a voice practice (for a debrief, with that result's feedback and your answers), a free practice answer and its question when you ask for AI feedback on it (not stored by Aethel), job description text a company pastes (not stored by Aethel), and in the live interview room your first name and, if you came from a company's link, its role details and questions.
    Location
    Not limited to one country. We use OpenAI’s standard API without its data residency option, so OpenAI can process requests in the United States or in other countries where it and its cloud providers run (its list names more than 20). For requests Aethel sends to OpenAI directly (such as scoring and written feedback), OpenAI keeps request content for up to 30 days in abuse monitoring logs. In the live interview room, the interviewer’s questions come from OpenAI’s GPT-4.1, and OpenAI’s GPT-4.1 mini checks whether you have finished speaking, both through LiveKit Inference, where neither LiveKit nor OpenAI keeps the content. If GPT-4.1 at OpenAI is slow to start a reply, the same model run by Microsoft Azure can answer instead (see its entry).Checked October 6, 2026. Sources: OpenAI subprocessors, OpenAI data controls, LiveKit Inference.
  • Microsoft Azure

    What they do for us
    In the live interview room, a backup for the AI model that writes what the interviewer says. If OpenAI is slow to start a reply, we also send the same request to the same OpenAI model (GPT-4.1) run by Microsoft Azure, and use whichever answers first. Only for interviews you agreed to from October 20, 2026.
    What they receive
    Only when OpenAI is slow to start a reply: the same request the OpenAI model gets. That is the interviewer’s instructions (with the profile details it uses to choose questions, your first name, and a company’s role details and questions if you came from its link) and the text of the interview so far. Never your audio, your video, or your camera.
    Location
    Not limited to one country. The request reaches Microsoft through LiveKit Inference, which by default can send a request to a provider in any region. In our tests in October 2026, LiveKit served every backup request from deployments it labels as Microsoft’s East US and East US 2 regions, both in Virginia, United States. Neither LiveKit nor Microsoft publishes whether those deployments keep the processing in the United States, so we do not promise a country. LiveKit says neither it nor its model providers log, store, or train on what we send, and that its contracts with them require this. Microsoft says requests to the OpenAI models it runs are not shared with OpenAI and are not used to train models. By default, Microsoft’s terms let it store requests that its abuse checks flag, in the same geography as the deployment, so its staff can review them, unless the account holder is approved to switch this off. The account holder here is LiveKit, which does not publish whether it is.Checked October 20, 2026. Sources: LiveKit Inference, LiveKit subprocessors, LiveKit data processing addendum, Microsoft: data, privacy, and security for Azure OpenAI models, Azure regions.
  • Our video interview provider, currently Beyond Presence

    What they do for us
    Shows Nelly, the AI interviewer, as a realistic video avatar in the live interview room and speaks her questions, and shows her the same way in a coach debrief or voice practice with video.
    What they receive
    The voice of each question or reply (which can quote your profile or earlier answers) and standard connection details such as IP address and browser. It does not receive your camera or microphone.
    Location
    Stored in the EU, processed in several countries. Beyond Presence stores what it keeps in the EU, on Google Cloud in Belgium and the Netherlands. It makes the video on its nearest available server, in Germany, the Netherlands, the United States, India or Saudi Arabia. Our requests come from our interviewer service in the United States, but the region is not guaranteed: we have not set up the region pinning or zero data retention it offers enterprise customers.Checked October 6, 2026. Sources: Beyond Presence trust FAQ, Beyond Presence privacy.
  • LiveKit

    What they do for us
    Runs the video call when your interview happens in our live interview room, and the call of a coach debrief or voice practice with Nelly, and connects it to the speech services below.
    What they receive
    The call audio, including your microphone audio while your microphone is on (it turns on by itself when it is your turn, and you can mute it), the text of the conversation, the profile details the interviewer uses to choose its questions, and standard connection details. Aethel does not record the call.
    Location
    United States. LiveKit stores what it keeps about our project in its United States data region, and our interviewer service runs in its us-east region (Ashburn, Virginia). The call’s audio and video pass through the LiveKit server nearest each person, which can be outside the United States for someone joining from another country. LiveKit says it does not keep media, and we switch off LiveKit’s session recording for every interview.Checked October 6, 2026. Sources: LiveKit data regions, LiveKit agent regions, LiveKit media routing.
  • Deepgram

    What they do for us
    In the live interview room, and in a coach debrief or voice practice, turns what you say into text.
    What they receive
    Your microphone audio while the microphone is on. If you type your answers with the microphone off, it receives nothing from you.
    Location
    Not limited to one country. Your audio reaches Deepgram through LiveKit Inference. LiveKit lists Deepgram as a United States company, but LiveKit can send these requests to any region and does not publish which one serves our account. Nothing is kept: LiveKit requires Deepgram not to store, log or train on the audio.Checked October 6, 2026. Sources: LiveKit Inference, LiveKit subprocessors.
  • Cartesia

    What they do for us
    Creates Nelly’s voice in the live interview room, a coach debrief or a voice practice.
    What they receive
    The text of each question or reply Nelly speaks, in an interview, a debrief or a voice practice.
    Location
    Not limited to one country. The question text reaches Cartesia through LiveKit Inference. LiveKit lists Cartesia as a United States company, but LiveKit can send these requests to any region and does not publish which one serves our account. Nothing is kept: LiveKit requires Cartesia not to store, log or train on the text.Checked October 6, 2026. Sources: LiveKit Inference, LiveKit subprocessors.
  • SendGrid

    What they do for us
    Delivers account and interview emails, passes messages from our support form to our team inbox, and does the same for company requests.
    What they receive
    Your email address, your name, and the email content, including what you write in the support form or a company request.
    Location
    North America. We send email through SendGrid’s standard service (smtp.sendgrid.net), not its EU data residency option. Twilio, which runs SendGrid, lists the companies that host and store SendGrid email in North America, one of them also in the EU.Checked October 6, 2026. Sources: Twilio subprocessors, SendGrid EU data residency.
  • Sentry

    What they do for us
    Tells us when something breaks on our website or our server.
    What they receive
    Only when an error happens: technical details, such as the error itself, the page address with private link codes removed, your browser type, and the recent steps on the page before the error, such as pages opened and buttons pressed. We remove email addresses and private codes before a report is sent, and we do not attach your name or account. Nothing is sent on a normal page visit.
    Location
    United States. Our website and our server send error reports to Sentry’s US region (ingest.us.sentry.io), which Sentry stores in Iowa.Checked October 6, 2026. Sources: Sentry data storage location.
  • Stripe

    What they do for us
    Processes payments from companies.
    What they receive
    Billing email and payment details. You enter card details on Stripe; Aethel never sees full card numbers.
    Location
    United States. Stripe names Amazon Web Services and Microsoft Azure in the United States as the cloud providers for its service. Its privacy policy says it may also transfer data to other countries, including India.Checked October 6, 2026. Sources: Stripe service providers, Stripe privacy policy.
  • Anthropic (Claude)

    What they do for us
    Aethel's AI support assistant. It reads support messages, and for a review request that interview's transcript and scores, to write the reply.
    What they receive
    What you write to us in a support message, a company request or a review request, with your name and email address, and for a review request, that interview's transcript and scores. Nothing from an interview you have not asked us about.
    Location
    United States and other countries. Anthropic’s privacy policy says it moves personal data to its servers in the United States, and may also move it to other countries. Anthropic keeps and uses these conversations under its consumer terms and privacy policy. The Claude account Aethel uses for support has Anthropic’s model training setting turned off, so Anthropic does not train its models on them, except on a conversation flagged for a safety review or sent to it as feedback (we send none as feedback). With training off, Anthropic keeps them for 30 days, or for up to 2 years if its safety systems flag one, and longer if the law requires it or to settle a dispute.Checked October 7, 2026. Sources: Anthropic privacy policy, Anthropic consumer terms, Anthropic: how long it keeps data, Claude Code data usage.

The 3-minute voice try on the free practice page uses four of these: LiveKit (the call), Deepgram (your voice to text), OpenAI (what Nelly says) and Cartesia (Nelly's voice). See the privacy policy.

Location is where each company stores or processes the data it gets from Aethel. We checked it against each company's own published documents and our own account settings, on the date shown. Where a company does not publish one country for our account, we say so instead of guessing. If you need more detail, email support@aethel.pro and ask.

Coach debrief and voice practice

The coach debrief and the voice practice on your results page are spoken conversations with Nelly, Aethel’s AI interviewer, here acting as a coach. Your microphone turns on when one starts, so Nelly can hear you. They run through LiveKit (the call), Deepgram (turns your voice into text), OpenAI (writes what Nelly says; in a debrief it also reads that result’s feedback and your answers) and Cartesia (Nelly’s voice). When there is video, Beyond Presence makes Nelly’s video from her voice. Microsoft Azure is not used for them.

Aethel keeps no recording and no transcript of a debrief or a practice. For a debrief it keeps when it started and ended and how many minutes it took, with that interview result, and deletes it with the result: up to 24 months after the interview, or sooner if you delete that attempt. For a practice it keeps when it started and ended, how many minutes it took, and how many seconds of Nelly’s video it used, if any, for up to 24 months. With each, it also keeps usage numbers for its own costs (seconds of video, characters spoken by Nelly, model tokens), never your words. Deleting your account deletes both at once.

Other services that receive limited data

  • Your browser's speech recognition: in the backup interview, used only if the live interview room cannot start, your browser starts listening by itself after it reads each question aloud and turns your speech into text. On the free practice page your browser listens only when you press "Speak your answer", and turns your speech into text there; Aethel never receives that audio. In some browsers, such as Chrome, the browser maker processes that audio under its own terms. In the live interview room, the coach debrief and the voice practice, Deepgram does this instead (listed above). You can always type instead.
  • Your browser's voice: in the backup interview, which has no video, your browser can read each question aloud. Aethel uses only voices that run on your device, so the question text is not sent anywhere to be spoken.

Former providers

These companies processed data for Aethel before, and no longer do. What they received before the date shown stays with them under their own terms. If you want us to ask one to delete it, email support@aethel.pro.

  • D-ID, until October 7, 2026. It showed the AI interviewer as a video avatar in the backup interview, and was the backup video in the live interview room. What it received: The text or voice of each question (which can quote your profile or earlier answers) and standard connection details such as IP address and browser. In the backup interview its software ran in your browser and stored a random ID there. We switched off its usage analytics with the setting D-ID provides; if it still sent usage events, they went to Mixpanel, an analytics service D-ID uses. D-ID stores data on Amazon Web Services in Oregon, United States, and deletes data from its live agents within 72 hours, but its terms let it keep using what it received. Source: D-ID privacy policy.

About the AI providers

We send interview content to OpenAI through its API. Under OpenAI's API terms, content sent through the API is not used to train OpenAI's models by default. Aethel does not use your interview to train AI models, and we do not share it with anyone to train theirs. If OpenAI is slow to start the interviewer’s next reply, we also send the same request to the same OpenAI model run by Microsoft Azure, through LiveKit, and use whichever answers first. This applies only to interviews you agreed to from October 20, 2026, when the consent screen started naming Microsoft Azure. When the interviewer’s request also goes to Microsoft Azure (only if OpenAI is slow), Microsoft says it does not share what it receives with OpenAI or use it to train models, and LiveKit says neither it nor the model providers it uses log, store, or train on it. If you send a support message or ask for a review, Aethel's AI support assistant (Claude, made by Anthropic) reads it, and for a review, that interview's transcript and scores. The Claude account we use for support has Anthropic's model training setting turned off (checked October 7, 2026), so Anthropic does not use those conversations to train its models. Anthropic's terms make two exceptions: a conversation flagged for a safety review, or one sent to Anthropic as feedback. We do not send support conversations as feedback.

How we tell you about changes

  • When we add, remove, or change a company on this list, we update this page and our privacy policy, and change the "last updated" date at the top.
  • We do not send email notices about changes to this list today.
  • If your company needs notice before a change for contract reasons, email support@aethel.pro and tell us.

The same list, with more detail on what we collect and how long we keep it, is in our privacy policy.