Privacy Policy

Crush Job is an AI product, and AI is not something your own computer does by itself. To tailor a resume or transcribe an interview, the app has to send your text or your audio to an AI provider. That is the plain fact this policy starts with, because it is the one that matters most and the one a privacy policy is most tempted to bury.

What is true is narrower, and we think still worth choosing us for: with Cloud Data Sync off — the default — we store none of your content on our servers, your files and history live on your own machine, and the AI work we route on your behalf is forwarded to the provider without being kept.

For the short version in one table, see where your data goes.

What this policy covers

This policy covers both the crushjob.ai website and account (sign-up, billing, and support) and what the desktop app sends when you use it. crushjob.ai and the Crush Job desktop suite are operated by Crush Job LLC, a Washington limited liability company.

What the desktop app sends, and to whom

The app runs on your machine and keeps your profiles, resumes, and usage history there. Doing the AI work, however, means sending the relevant part of your data to the provider that performs it:

• Anthropic — Resume text, the job posting, LinkedIn profile text you paste, the answers you type into gap and strengthen questions, and interview transcripts — the words transcribed in Crush Coach (yours and the other participant's) and in Mock It! — sent to be tailored, scored, and turned into battle cards and interview questions, and to write Crush Coach's live suggestions and debriefs and Mock It!'s follow-up questions and scores.
• AssemblyAI — Live audio, when you use Crush Coach or Mock It!, for real-time transcription. Crush Coach captures two streams: your microphone and your computer's audio output, which contains the other participant's voice.
• Deepgram — In Mock It!, the interviewer's question text, which Deepgram turns into speech with its Aura-2 voice. If Mock It! is set to use Deepgram for transcription, your microphone audio is also sent to Deepgram for real-time transcription. AssemblyAI is the default.
• Crush Job sends one anonymous message the first time it opens, so we know the download worked. It contains a random ID, your operating system, and the app version. The operating system includes its build number; the message may also include a download reference. We use your IP address during the request to look for a matching download, but do not store it with the install record. Opening the account page from the app can also match the install to your browser's download cookie. Linking a device connects this record to your account. We delete unlinked install records after 400 days and erase linked install records when your account is purged.

Text requests go through crushjob.ai. The app sends them to us, we forward them to Anthropic on our own provider account, and we bill the usage against your plan's credits. Audio works differently: our server gives the app a pass that expires within a minute, and the app uses it to stream your audio straight to AssemblyAI (or to Deepgram, if Mock It! is set to use it for transcription), and, in Mock It!, to send the interviewer's question text straight to Deepgram to be spoken aloud — the audio does not pass through our servers. That is how the suite works for every linked account — there is no provider key for you to obtain, and the app has nowhere to enter one. Routed, not kept: we pass each text request to the provider without keeping its text. For each text request we record the model name, the token counts, the operation label, your app version, the provider's stop reason and the credits it cost; for each audio session, the provider, the audio seconds or characters and the credits it cost — never your prompts, resumes, or transcripts. Those usage records are kept until you delete your account. For a few operations we also keep a one-way fingerprint (a SHA-256 hash) of the AI's reply, so we can check that a retry is resending a reply we really sent; it cannot be turned back into the text, stops being used after 30 minutes, and is deleted by a daily sweep after that.

Crush Coach records the other side of a conversation. It captures your computer's audio output, which contains the other participant's voice, and sends it for transcription. Recording laws vary by jurisdiction — in some places every participant must consent before a conversation is recorded or transcribed. Obtaining that consent is your responsibility, not ours.

Crush Coach asks before it records, and you can take that back. Before your first session, Crush Coach shows what it captures and where the audio goes, and starts only once you agree. To withdraw that agreement, open Crush Coach's Settings and choose Withdraw Recording Consent. Crush Coach then won't start another session until you agree again. A session already running is not stopped.

If you were the other person on that call. If you were on a call that a Crush Job user transcribed and want it removed, email support@crushjob.ai. The transcript and any recording are saved on that user's computer, not on our servers, so we cannot reach them. If that user turned on Cloud Data Sync, our servers hold a summary of the session — the company, when it started, whether a debrief exists and how many turns it had — but never the transcript or the audio. We will delete that summary when you ask. During the call, your voice went to the transcription provider and the transcribed words went to Anthropic; see what the AI providers keep below.

What the AI providers keep

How long each AI provider keeps what it receives, and whether it trains its models on it, is set by that provider's terms for API customers. Here is what each one's own page said when we checked it on October 3, 2026:

• Anthropic deletes API inputs and outputs within 30 days, unless it has to keep them longer, for example to enforce its usage policy or because the law requires it (Anthropic's retention page). By default, it does not use API inputs or outputs to train its models (Anthropic's training page).
• AssemblyAI keeps no streaming audio or transcripts for accounts that have opted out of model training, keeps some metadata for logging and billing, and uses only certain files, as its contract allows, for training (AssemblyAI's data retention page).
• Deepgram keeps audio and text to improve its models by default, unless a request opts out; for a request that opts out, it does not store the audio, text or speech after replying (Deepgram's data page). Mock It!'s interviewer voice requests opt out, from app version 5.14.0. Live transcription is handled by AssemblyAI, not Deepgram.

Where your files are kept on your computer

The app keeps its own data — your settings, person profiles, usage history and this device's link to your account — in a Crush Job data folder:

• Windows: %LOCALAPPDATA%\CrushJob
• macOS: ~/Library/Application Support/CrushJob
• Linux: ~/.local/share/CrushJob, or a CrushJob folder inside $XDG_DATA_HOME if you have set it

Your documents are kept apart from that, in the folder you chose for your master resume. Tailored resumes and cover letters go in its Applications folder. Battle cards, interview transcripts, debriefs and recordings go in its Interviews folder.

Uninstalling does not delete your documents. On Windows, uninstalling from Settings asks whether to keep your settings and data, and keeping them is the default. If you choose to remove them, the uninstaller deletes the Crush Job data folder above, and nothing else. An uninstall run without that question, such as by a management tool, keeps it. On macOS and Linux, removing the app leaves both folders where they are. To remove everything, delete those folders yourself.

How we protect it

• Everything the desktop app sends to us and to the AI providers travels over encrypted connections (HTTPS/TLS), including the live audio streams. Our website tells your browser to use only encrypted connections.
• The pass that lets the app stream audio to a provider expires within a minute of being issued.
• Your account data is stored with Supabase, which encrypts all customer data at rest with AES-256 (Supabase security, checked October 3, 2026).
• On your computer, the link between this device and your account is stored encrypted by your operating system: Windows data protection, the macOS Keychain, or the Linux Secret Service. On macOS and Linux, the app also limits its data folder so only your user account can open it.

Cloud Data Sync

Cloud Data Sync is off by default. One switch in the Hub's Account tab covers every profile on that computer — it is not set per profile. Nothing in this section applies to you unless you turn it on.

When it is on, the app pushes structured copies of your own work to our servers so it can show up on your crushjob.ai dashboard: the profiles, applications, battle cards, mock-interview sessions and Crush Coach session records you create in the app.

It is a copy, not a move. The documents on your own machine remain the original record — your resumes, cover letters, transcripts and recordings. Cloud Data Sync never deletes or edits any of them.

For a battle card, that includes the card itself as the app shows it — the snapshot, evidence map, STAR stories, likely questions and answer outlines — so you can read and download it from your dashboard. The practice answers you type while strengthening a card are not included.

Your Job Tracker entries work differently, because you can keep them on more than one computer. Change a job in the app on a second computer you have linked, and that change is sent back to the others. It can update the matching entry there, or remove it if that is what you did. The most recent edit wins, so a newer change you made on one computer is not overwritten by an older one from another. You can also add, change or delete a Job Tracker entry on crushjob.ai itself, and that change is sent to each linked computer the same way.

Turning Cloud Data Sync off stops both directions at once. Nothing further is sent, and nothing further is received.

Your interview transcripts and audio recordings are not part of it. A mock interview and a Crush Coach session each leave a full turn-by-turn transcript on your machine, and Mock It! can keep an audio recording as well. What Cloud Data Sync uploads is the structured summary beside them — the company and role, the score, and for a Crush Coach session whether a debrief exists and how many turns it had. The transcript, the debrief text, the coaching feedback and the recording stay where they are.

While the app is open it syncs about every fifteen minutes, and you can press Sync Now at any time. Closing the app also sends one last sync of whatever changed on that computer, for up to a few seconds; anything that does not finish goes the next time the app opens.

You can delete the cloud copy at any time — all of it at once from the Hub's Account tab, or all of it or one profile at a time from the Cloud Data Sync page of your account on crushjob.ai — and deleting your account removes it as well. We do not share synced data with third parties, and we do not use it to train AI models.

Crash and error diagnostics

If your device is linked to an account, the desktop app sends us a small report when it hits a crash or an unhandled error: the exception type, the error message, where in the code it happened (the names of the modules and functions involved, and line numbers), the app version, and which operating system build you run it on (Windows, macOS or Linux). Any text in that report is scrubbed on your device before it is sent: home-folder and other user-folder paths, email addresses, long number sequences, web addresses and document file names are removed.

These reports are about the crash, not about you: they never include your resume text, interview transcripts, prompts, or any file contents. That makes them metadata, not content, so this does not change what we say above — unless you turn on Cloud Data Sync, your content is never written to our servers. The desktop app's crash reports go to our own servers, the same ones the rest of crushjob.ai runs on. When a new kind of crash arrives, or a fixed one comes back, we email its details to our support inbox through Zoho ZeptoMail and post them to the task board we use to track fixes. That email and that task-board entry carry the report above, how often it has happened and on which app versions, and the email address of each account that reported it. The website is a separate case, and the paragraphs after this one say how.

This is not a setting you can turn off on a linked device — we rely on crash reports from every linked install to find and fix issues across the app. If your install is not linked to an account, no crash reports are sent at all.

The website reports its errors to Sentry. When a page on crushjob.ai hits an error in your browser, or one of our own endpoints fails while serving you, we send a report to Sentry, an error-monitoring service acting as our processor. That report carries the error and where in our own code it happened, the page you were on, your browser and operating system, which build of the site you were running, and a short trail of the actions and requests that led up to it.

If you are signed in, the report also carries your account's id — the same identifier your account already has in our database — so we can tell one person hitting a problem ten times from ten people hitting it once. It does not carry your name or your email address. We do not record or replay your screen: Sentry offers that and we have it switched off. For errors raised on our servers rather than in your browser, we tell it to leave out request bodies and headers.

We also send Sentry a small sample of performance traces — about one in ten page loads and requests — so we can find the pages that are slow. A trace records which page or endpoint was used, how long each step took and whether it succeeded, and, if you are signed in, the same account id as above. When a trace comes from our servers, we remove request bodies, headers and your IP address before it leaves them, and we have told Sentry not to store IP addresses from any report.

Information we collect

• Account data — your email address, and (if you sign in with a third-party account — Google, Microsoft, LinkedIn, GitHub, or Apple) the name and email address that provider shares with us, to create and secure your account. If you sign in with Apple and choose Hide My Email, we only ever see the private relay address Apple generates, never your real one.
• Contact data — the name, email address and message you submit through our contact form, and the IP address it was sent from. We keep these with no set deletion date.
• Billing data — handled by our payment processor (Stripe). We store a customer reference, your subscription status, and what Stripe reports about each payment — the amount, the card brand and its last four digits, and links to the receipt and invoice — never your full card number. We also keep a fingerprint Stripe generates for the card you paid with — a code that recognises the same card again, not the card number — used only to stop one card claiming a referral reward twice.
• Browser mock-interview demo — on /try-mock-interview you can type answers to three interview questions without an account. The role you pick and your answers are sent through our server to Anthropic, which scores them. We do not save the answers or the scorecard. To limit abuse we count requests against your IP address for up to an hour, and we keep a daily total of how many scorings ran, with nothing about the visitor. Cloudflare Turnstile checks the form for bots.
• Download records — when you download the installer, which one, your account, the IP address, your browser user-agent, and the campaign tags and ad click ID on the link you came from. We keep these with no set deletion date; deleting your account removes the link to it and keeps the rest.
• Product analytics — usage of the website (for example, which pages convert, and account-linked events like sign-up or a completed download) collected via PostHog, to improve the site. This only runs if you accept the cookie banner; declining blocks it entirely.
• Email engagement — what happened to a message we sent you: whether it was delivered, bounced, had a link followed (in our lifecycle emails only), or led you to unsubscribe. We do not record whether a message was opened. See the email section below for what is and is not recorded.

Cookies

We use a small number of cookies: session cookies required to keep you signed in, and — only if you accept the cookie banner shown on your first visit — analytics cookies from PostHog and ad-measurement cookies from Google Ads to see which ad led to a visit or signup. We do not use Google Ads for personalized ads or audience building. You can change your mind at any time via the "Cookie preferences" link in the footer. We also use Cloudflare Turnstile, a bot-detection check on our forms, which does not track you across sites and runs regardless of your analytics choice since it's required to keep the forms working. One more cookie remembers, on this browser only, which sign-in method you used last — so the login page can point you back to it. That cookie holds the method's name only, never your email address. Two short-lived cookies (five minutes and thirty minutes) mark that a sign-up or sign-in just happened, so we can record that single event — they hold only an event name and a coarse method label, never an email address or a token, and expire on their own shortly after.

If you arrive from a marketing link (for example, a post on LinkedIn or Reddit with tracking parameters in the URL), we set one further first-party cookie recording which link it was, for up to thirty days. If you create an account within that window, we record that attribution on your account so we know which channel it came from — this runs regardless of your analytics/cookie-banner choice above, is never shared with any third party, and is not used to track you across other sites. It exists purely so we can tell which of our own posts and listings are worth the effort of continuing.

A first-party cookie also records which ad or link brought you here, including its click ID, and is kept for ninety days regardless of your cookie-banner choice.

When an ad click leads to a confirmed signup, installer download, or emailed score report, we record the click ID and the time of the action for ad measurement. These records do not contain your email address or resume score. Signup and download records can link to your account.

What we do not collect

• AI provider API keys — there is no key of yours for us to collect. The suite reaches the providers listed above on our own provider accounts, and the app has no field for entering a key.
• Your resumes, prompts, or interview transcripts — with Cloud Data Sync off, we do not store them. They do pass through our servers on the way to the provider that performs the work, and are not retained there. If you turn Cloud Data Sync on, structured copies of your applications and interview sessions are stored so they can appear on your dashboard — see Cloud Data Sync above. Either way, the provider that performs the work does receive them — see what the desktop app sends above, and that provider's own policy for what they do with it.

How we use your information

• To provide the website, your account, and billing.
• To send you email about your account — the acknowledgement our contact form sends back, purchase receipts, credit-balance notices, and changes to our terms.
• To send you email you can switch off — recaps of your own activity, product updates, and job-hunt tips or offers. See the email section below.
• To understand and improve how the site performs.

Email we send, and how to stop it

We send four kinds of email. They differ in what they are for, and in whether you can turn them off.

• Account — purchase receipts, credit-balance notices, changes to our terms, a one-time sign-in link we email you, and the acknowledgement our contact form sends back. These come with having an account, so they carry no unsubscribe link.
• Activity — recaps of your own use of the suite, such as a monthly summary or a milestone you passed.
• Product updates — what we have shipped, and what is coming next.
• Tips and offers — job-hunt advice, help getting started, and occasional promotions.

The last three are on by default, and you can turn any of them off whenever you like. Every one of those messages carries an unsubscribe link that takes effect in a single click, and you can set all three at once on your email preferences page. Choosing to stop everything is recorded against your email address rather than your account, so it still holds if you sign up again later — and unlike the individual switches, it stops account email too.

Two providers deliver this mail. Zoho ZeptoMail carries account and activity email; Amazon SES carries product updates and marketing, sent from a separate subdomain so that trouble with one kind of mail cannot affect delivery of the other. The site can also send that second kind through Resend instead, one provider at a time. Each of them receives your email address, the message we wrote, and what became of it — delivered, bounced, or reported as spam. None of them receives your resume text, your prompts, or your interview audio.

We count clicks in our lifecycle emails, and nothing else. Links in our lifecycle emails go through crushjob.ai so we can count clicks. Those are the getting-started series, the monthly summary, milestone notes and the win-back email. A link goes to crushjob.ai first and then forwards you straight to the page it names. We use no tracking pixels, so no email contains an invisible image and we do not know whether you opened a message. The counting runs on our own servers — no advertising network or analytics vendor is involved.

We keep those records deliberately thin. A click is stored without your IP address and without your browser user-agent, and our database rejects a record that carries either. We do glance at the user-agent for an instant, only to tell a person from a mail-security scanner that follows links on your behalf, and we do not keep it. We record that a link was followed, not where you were when you followed it. We use the counts to see which emails help, and we compare the time of a click with when your desktop app last reported in, using data we already collect; that comparison is reported only as totals.

Transactional email is not tracked. A receipt, a low-balance notice, a password or security message, or a change to our terms reaches you with no tracking image and no rewritten links, because that mail is not something we measure. No email we send contains a tracking pixel.

Service providers

To run the website and your account we use Supabase (database and auth), Stripe (payments), Zoho ZeptoMail (account and activity email), Amazon SES or Resend (product-update and marketing email), PostHog (product analytics, opt-in), Google Ads (ad measurement, only after you accept the cookie banner), and Cloudflare Turnstile (bot protection on our forms). Each processes data only to provide their part of the service.

These are separate from the AI providers named above, which are the ones that receive your actual content. Both lists matter; only one of them is about your resume. The email providers sit on this list rather than that one for exactly that reason: they are handed an address and a message we composed, never anything you wrote into the app.

Signing in with a third-party account — Google, Microsoft, LinkedIn, GitHub, or Apple — is also governed by that provider's own privacy policy. We receive only what the sign-in itself shares with us: your name and email address, nothing more.

Your rights

You can request access to, correction of, or deletion of your account data at any time by emailing support@crushjob.ai.

You can also delete your account yourself, from the bottom of your account page on crushjob.ai. That deletes the account and how you sign in to it, your plan and credit balance, your linked devices, and everything Cloud Data Sync stored. It can't be undone. A plan that still renews has to be cancelled first, so you aren't charged again. Records of your payments stay with Stripe. We keep a short note that the account was deleted — its email address, your name and the date you joined — so we can answer questions about it, and our record of each email we sent you (the address, which email it was, and whether it was delivered) stays too. Neither has a set deletion date.

On your computer, you control what is sent. The AI features send your content only when you start one, such as tailoring a resume or building a battle card. Crush Coach and Mock It! send audio for transcription only during a session you start. Cloud Data Sync is off unless you turn it on.

You can stop sharing at any time. Stop starting AI features and sessions, and nothing more is sent to the AI providers. To withdraw your recording consent for Crush Coach, use Withdraw Recording Consent in Crush Coach's Settings. Delete the folders described in where your files are kept to remove your data from your computer, and delete your account as described above to remove what we hold. You can also email support@crushjob.ai to withdraw your consent and ask us to delete your account data.

Please don't put health, financial or government ID details into your resume or your answers. The app does not need them, and anything you include is sent to the AI providers along with the rest.

Contact

Questions about this policy? Email support@crushjob.ai. Crush Job LLC is based in Renton, Washington.