What is held about you
Version 2026-11-02. This is the version your account records having accepted, alongside the terms.
Written by the person who built the service, accurate about what the software does, and not reviewed by a lawyer. Still to come, and named so the gap is not silent: the operator's full name and service address in a legal notice, a signed data processing agreement with OpenAI, a decision on whether to ask Anthropic and OpenAI to keep nothing at all of what is sent, a formal record of processing, and a breach procedure. Until those exist, treat this as a truthful description rather than a finished legal document.
Who is responsible
One person, operating from Germany, who also administers the server. There is no company, no data protection officer and no team. The name and a service address belong in a legal notice that does not exist yet; until it does, the way to reach the operator is the address on the support page.
What is held, and why
| Your account | Email address, a scrypt hash of your password (never the password), a display name and handle, your language and time zone, and the version and date of the terms you accepted. Sign-in timestamps. Needed to run the service you asked for. |
|---|---|
| Your CV | The file you upload, the text extracted from it, and the structured facts read out of it — employment history, education, projects, certifications and licences, skills and languages, plus your name, email address, phone number, location, links, a one-line headline, citizenship and work authorisation where your CV carries them. No date of birth is read out of a CV; if yours shows one, it is in the file and its text like everything else on it. From a CV uploaded as a YAML file, the personal details are kept as you typed them. If you change your CV on this site, or build one here, the draft you work on and each version you publish are held too: what was carried over from the CV it started from, if it started from one, and what you typed, kept as you typed it apart from spacing and the bullet marks at the start of a line, with a note on each line under a job, project, study or certificate of whether it came from a CV file of yours or was typed here. The citizenship and work authorisation read out of an upload are never carried over into it, except a right-to-work line you wrote yourself in a CV uploaded as a YAML file. It also holds the language your CV is written in: the page suggests one when it can tell from your CV's own words, which are read for it on this server, and holds the one you choose, which can be "Another language". It has no place for a date of birth, a nationality or a marital status; if you type one into it after its label ("Nationality: …") in English, German, Turkish or Swedish, anywhere but your name, email, phone and links, the page points it out so you can take it out, but it does not find every way of writing one. This is the most sensitive thing here, and it is what the whole service works from. |
| What you are looking for | Countries you would move to and countries you would not, your seniority band, how old a posting may be, a home city and a radius around it, and whether you would relocate; the words in a job title, and the employers, whose postings you never want to see; the job sources you hide from your queue and how many postings it shows at a time; the languages you said a CV or cover letter for you may be written in; and the countries you said you can already work in without a visa sponsor. |
| Judgements | For every posting in the shared collection, whether it passed your filters and the reason it did not. This is the largest thing held about you by row count — tens of thousands of rows per account — and it is entirely derived from the two rows above. |
| Applications | Which postings you marked as sent, when, the outcome you recorded, and any note you wrote to yourself. |
| Jobs you added by link | Which job you added and when: one record per job. For a job on one of the application systems the service reads, or on the Bundesagentur für Arbeit's job search, nothing else. Not the link as you pasted it: the job is read from that system, and the link kept with it is one we rebuild from its parts, or that system's own. The posting itself is the employer's, like every other in the collection. That you added it is yours, which is why it is in your download. |
| Jobs you added from a page | For a job read from any other web page, also: the page's address, as the job's link; the job's fields as you saved them, and which of them you chose from the page and which you typed; and the first 200 characters of each paragraph you left out, for the person who checks the job. Under Adding a job from any other page, below, is the whole list. The page's text goes when the job closes here (under How long). |
| Reports | Which job you reported as not genuine, and when. There is no box to write in, so there is nothing else. |
| Notices | The notices on your account page about an employer page you added a job from or applied to, or about a job you added from a page or from the Bundesagentur für Arbeit, or applied to: what kind (approved, kept private, declined, found not genuine, removed on request, taken down, hidden after reports, or published from somebody else's copy), which page and which of your jobs, or the job we have in its place, when, and when you dismissed it. No wording is stored; it is written each time the page is shown, and a declined job's reason is read from the job's own record. |
| Add-a-link attempts | Each time you paste a link that does not take you straight to a job you can already see here: that your account made an attempt, and when. Not the link and not its website. Kept for 1 hour, to slow down a flood of links from one account. |
| Pages read for you, and refused links | Each time this server asked a site anything to read a link you pasted as a web page, and each time one of those links was refused for where it pointed (a private network address, say, or too many redirects): that your account did, and when. Not the link and not its website. Kept for 24 hours, to limit how many pages are read for one account in a day, and to pause adding from pages after repeated refusals. |
| Sessions | A hash of your sign-in cookie and its timestamps — not the cookie itself, so a copy of the database cannot be used to sign in as you. No IP address and no access log. |
| Messages sent to you | Which messages were sent to you, when, and what the mail provider said about each. Deliberately not a second copy of your address. |
| Emails to support | What you write to the address on the support page: your email address, your message and anything attached, and the replies sent back. Not in this service's database; they are in a mailbox at Zoho, described below. |
| Sign-up and reset attempts | The network address an attempt came from and which action it was, kept for one hour, to stop someone hammering the sign-up form. No email address is recorded here — the address attempted is the thing being protected, not the thing being limited. |
The lawful bases are: performing the service you asked for (account, CV, preferences, judgements, applications, the jobs you add and the pages read for them, your reports, the notices and warnings about them, and answering what you write to support); a legitimate interest in keeping the service standing and its postings genuine (sessions, delivery log, rate limits, the limits on adding links and reading pages, the count of websites, error reports, the review of employer pages and of the jobs you add, and the operator's list of accounts); and your consent where you have given it. People named in a job ad are under their own heading, below.
Who else can see it — including the part that is uncomfortable
The operator can read everything. This is one database and one directory tree on one server, with no encryption at rest and no per-account separation on disk. The application itself keeps accounts strictly apart — no page accepts an account identifier, no screen shows one person's CV or decisions to another, and there is no "view as" — but anyone with shell access to the machine reads all of it, and the operator is that person by definition. That is a property of running this cheaply on one box, not an oversight, and you are entitled to know it before you upload a CV. The only screens that show anything about one account to another are the operator's own reviews of the employer pages, and of the single jobs, that somebody added by link, the operator's list of accounts, one notice that can tell you that somebody else added the same page as you, and some of the add page's answers to a link you paste, each never who. Only the operator's account can open the reviews and the list. The reviews, the notice and the answers are described under Jobs you add by link below, and the list next.
The operator's list of accounts. A page only the operator's account can open lists every account on this service, yours included, with: its handle and display name; whether it is the operator's account or a member's; where it stands, which is one of: waiting for a CV, reading a CV, checking a CV it read, a CV that could not be read, being prepared, active, disabled, or being deleted; whether its email address has been confirmed, not the address; whether it has a confirmed CV, not the CV; how many postings have passed its filters and are waiting for its decision, which can include some that have since closed; how many postings have been judged for it, as a number: not which postings, and not what was decided; when its queue was last ranked, and whether that pass failed; when it last used the site, to within five minutes; the day it was created. It shows counts and states only: not which postings are in your queue, not what was decided about any of them, and not your CV or your email address. Nothing new is recorded for it: it counts and reads records this service already keeps to run your account. It is there so the operator can see that every account's postings are being judged and ranked, notice an account that is stuck, and answer what you write to support.
Anthropic (United States) receives the text of your CV when you upload it as a PDF or a Word file. A CV uploaded as a YAML file is not sent, and nothing else is: postings are judged and ranked on this server, and no application is drafted by a model. Nothing you type or change in your CV on this site is sent either. From a PDF or Word file, the whole text read out of it is sent, with your name, your contact details and everything else on the CV, and what comes back is the structured version you are shown to check. Since this version of the notice, Anthropic is asked first; before it, the text went to OpenAI alone (below). Anthropic deletes what was sent and what came back within 30 days. It keeps them for up to two years if its safety checks flag them as breaking its usage policy, and the scores those checks give for up to seven years, and longer only where the law requires it. It does not train its models on what is sent to its API. The account is held under Anthropic's commercial terms, which include its data processing addendum.
OpenAI (United States) receives the same text whenever Anthropic is not asked, or fails, refuses, cuts its answer off or cannot be reached: the same request then goes to OpenAI, so on such an upload OpenAI receives your CV instead of Anthropic or as well as it. From 10 September 2026 until this version of the notice, OpenAI received every such upload. Every request tells OpenAI not to store it. OpenAI still keeps its abuse-monitoring logs, which can hold what was sent and what came back, for up to 30 days, longer only where the law requires it or to protect against harm, and it does not train its models on what is sent to its API. Until the version of this notice dated 10 October 2026, requests did not tell OpenAI not to store them, and OpenAI kept those for 30 days too. No data processing agreement with OpenAI has been signed yet.
netcup GmbH (Germany) hosts the server, and the machine is in their Vienna, Austria data centre — inside the EEA. An earlier version of this notice said Nuremberg. That was the plan; Vienna is the machine.
Resend (United States) sends every email the service sends by itself: address verification, password resets, a note when somebody tries to sign up with your address, the confirmation when you ask to delete your account or cancel that, and the warning when an employer page, or a job, you added or applied to is found not genuine or taken down (below). So it receives your email address and what those messages say. Their delivery logs and message metadata are held in the United States whichever region mail is dispatched from. An earlier version of this notice promised a provider with data residency in the EU. That promise was not kept, and correcting it is better than quietly dropping it. Resend's records, and what Anthropic and OpenAI keep of the CVs they are sent (above), are the only parts of this service kept outside the EEA today.
Zoho (Zoho Corporation GmbH, Germany, whose EU data centres are run by Zoho Corporation B.V. in the Netherlands) hosts the support mailbox, the address on the support page, in Zoho's EU data centres. Whatever you send there is kept in that mailbox: your email address, your message and anything attached to it, and the replies sent back from it. Zoho says that mail in its EU data centres is stored in the EU, and that its staff in India may reach it to give technical support. The mailbox is not part of this service's software, which has no access to it: nothing in the software copies it into your download or deletes it, on a timer or when you delete your account. The operator reads it. A message stays there until the operator deletes it, or until Zoho does if it filed the message as spam, and Zoho keeps it for a while after that (under How long, below). To have yours deleted, or to get a copy, write to the same address and say so.
Backblaze stores the nightly backups, in a bucket in their Amsterdam region. What they hold is encrypted on the server before it is uploaded, with a key that exists only on the operator's own computer and has never been on the server or given to Backblaze — so they store a file they cannot read. This is the only copy of the database that exists off the machine.
Sentry (Functional Software, Inc., United States) receives a report when something fails on the server, and keeps it in their EU region, in Frankfurt, Germany. Only the server sends to Sentry: no page here loads anything from them, and every page tells your browser not to fetch anything from any server but this one. A report is the shape of the failure and not the person it happened to: the kind of error but not its message; where in this service's program it happened, with the few lines of the program around each step; for a failure the service's own code logs, the fixed wording of that line without the values filled into it; if it happened while answering a request, that request's method and which page it was for, by the page's pattern rather than its address (/job/{job_id}, never the job's number), or the method alone if it failed before reaching a page; and the time, the server's name, the versions of the software running, and a random identifier that belongs to that report alone. A report does not contain what you sent, your cookies, your IP address, which account was signed in, the values the program was working with, or the steps that led up to the failure, and nothing is measured about how long anything takes. Nothing at all is sent to Sentry while nothing fails, not even a count of the requests the site answered. Sentry keeps a report for 30 days on its free plan, which is the one this service uses, or for 90 days if it arrived while the account was on a trial of a paid plan, and after that it can no longer be opened.
Nobody else, and nobody from your browser. No analytics, no advertising, and no fonts, scripts or stylesheets loaded from anyone else's server — every page here is served whole from this machine, so browsing the service tells no third party what you are looking at. The server sends Sentry nothing while nothing fails; when a page fails, its report (above) names the page by its pattern, never by the number of a job, an application or a notice in its address. Links to a posting, or to where it says to apply, are opened with the referrer suppressed, so the site they lead to, an employer's or a job board's, is not told that you found it here. Adding a job by its link is different: it makes this server contact the employer's application system, or the site the link is on, at that moment, as described below, with nothing about you in the request.
Jobs you add by link, and reports
Adding a job from an application system. When you paste a link to a job on one of the application systems the add page names, or on the Bundesagentur für Arbeit's job search, this server, not your browser, asks that system for that posting, at an address built from the parts of your link; for such a link, the page it points to is never opened. On most of those systems that means reading the employer's list of open postings and keeping only the one you named; the Bundesagentur's posting is read from its own API, by the job's number. The request carries nothing about you, and nothing else from that list is kept. If the link is to a job we already have and you can see it, you are usually taken straight to it, with nothing recorded and nothing asked of the employer's system.
Adding a job from any other page. When you add a job from a link that is not on one of those systems, this server opens that page, first checking the site's robots.txt, the file where a site says what automated readers may open, and it does not open a page that file asks it to leave alone. If the page forwards to another address, it follows at most 3 such steps, and checks each new site's robots.txt too: at most 8 requests in all. LinkedIn, Indeed, Glassdoor and Seek are never opened. The requests carry nothing about you: no cookies, no referrer, nothing from your browser. They do tell the site that a get-me-hired.com user is looking at that posting at that moment, because our requests name us (our page for site owners says how). You then choose what the page says, and type only what it leaves out, never contact details. The page's text is held in this server's memory for up to 30 minutes while you fill in the form, and is not kept unless you save the job. If you do, what is kept is:
- the page's address, with the tracking and personal parameters we recognise removed, as the job's link;
- the job's fields, and which of them you chose from the page and which you typed;
- which paragraphs of the page you left out, by their first 200 characters, with the other choices the page offered, for the person who checks the job;
- how the page was read: the sites it forwarded through, whether the site had robots.txt rules, the words of its robots tag that we read (noindex, noarchive, nosnippet or none), and so whether it asked that its text not be shown elsewhere, a mark when a title or an employer's name of two words or more that you chose from it is also written in the link you pasted or in the page's address (as a search page repeats what was searched for), and a fingerprint of its text (a number computed from it, from which the text cannot be read back).
Not the page itself, and not the site's robots.txt.
Who sees it. A job on an employer page the service already reads (one on the operator's own list, one it found itself, or one the operator has approved) joins the shared collection at once, where everybody can see it. A job on a page nobody here has checked is visible only to the accounts that added it, and is marked not verified wherever it appears, until the operator has decided on that page. A job from the Bundesagentur für Arbeit is decided one job at a time, and is visible only to the accounts that added it until the operator has decided on it. If two people add the same unchecked job from an application system or from the Bundesagentur, both see it and neither is told that the other did.
Who sees a job read from a page. A job read from any other page is yours alone until it is checked: somebody else who adds the same page gets a job of their own. A person checks it before anyone else can see it, and can approve it only if the employer's own website vouches for the page: the page is on the employer's domain, or the employer's website links to that exact page or, when the page is on a platform that gives each of its users an address of its own (such as acme.teamtailor.com), to any page at that address. Once approved, the email addresses and phone numbers our scan finds in its title, its employer's name and its description are removed from the job, and other members see it, its description included; they see it without its description if the page asked that its text not be shown elsewhere (by the words noarchive, nosnippet or none in its robots tag). A job read from a link-sharing page (a Google Docs or SharePoint link, say) is never shown to any other member and never approved: nobody here confirms that it is genuine, though the person who runs this service can see it and may take it down if it is not.
The operator's check. Before deciding on an employer page, the operator can have this server look for a link from the employer's own website to that job board. The operator types the employer's domain from their own knowledge, never from the posting, and the server reads the front page of that site and at most 3 more pages on the same domain that look like its careers pages, searching them for a link to the board. The sites of the Bundesagentur für Arbeit, Arbeitnow and Arbetsförmedlingen, whose job feeds this service reads, are never read this way: such a domain is refused before anything is asked. Nor are LinkedIn, Indeed, Glassdoor and Seek, or this service's own site. The requests come from this server and carry nothing about you or about which job was added. What is kept, on the review of that page, is the domain typed, the address of each page and whether it could be read, and which job boards the site links to; not the pages themselves. The review also records facts about the board: how many postings it has, phrases common in scams found in them, and whether the employer is on a visa-sponsor register. None of it names an account that added a job there.
For a job read from a page, the same check looks for a link to that exact page instead (or, on a platform that gives each of its users an address of its own, to that address), and is skipped when the page is already on the domain the operator typed, unless it sits on a platform that hosts many people's pages or under a blog or upload path. It also asks how old the page's domain is, unless the page is on a platform that hosts many people's pages: this server reads IANA's list of domain registries, asks the registry of that domain when the domain was registered, and, if the registry does not say, asks crt.sh, a public log of web certificates, when the domain's first certificate was issued. Those requests name only the page's domain; like the rest of the check, they carry nothing about you or about which job was added. What they answered is kept on the review, with the phrases common in scams found in the job's text and the words around each, which are deleted with that text.
The decision, and what you are told. If the operator approves the page, its jobs become visible to everybody and the accounts that added one get a notice on their account page. If the operator finds it not genuine, or takes any employer page down later, its jobs are hidden from everybody, the people who added them included, and everybody who added or applied to one of its jobs gets a notice and, at a confirmed address, an email telling them not to send it documents. The email is fixed wording with a link to your account page; nothing written by whoever runs that employer page is put in it. Nobody who is warned is told who else was. A job from the Bundesagentur für Arbeit is decided the same way, one job at a time.
The decision on a job read from a page. The operator approves it (above), or decides one of the following, and you get a notice on your account page that says which, and why, and how to contest it:
- Kept private: it looks genuine, but the employer's website does not vouch for the page. It stays visible to you until it closes here.
- Declined: the posting has closed, the page is not one job, we already have the job, or it could not be verified. It is hidden, and its text is deleted at once.
- Found not genuine, or taken down after it was approved: it is hidden from everybody, and you, and anybody who applied to it, also get the warning email above. Its text is kept for 30 days, in case you contest the decision, and then deleted.
- Removed on request, when the site it was on, somebody who holds rights in it, or somebody it names asks: it is hidden, and its text is deleted at once. The notice says that this is nothing against you or the employer, and no email is sent. Such a request about a job already found not genuine or taken down changes nothing: that decision, its notice and the 30 days its text is kept all stand.
If the same job is approved from somebody else's copy of the page, yours stays visible to you until it closes here, and its notice says so. Only a job found not genuine or taken down sends the warning email; when several of yours are found not genuine at once, you get one email. The operator can also take one paragraph out of a job's saved text, without taking the job down.
Reports. Every job page has a report this job button. A report records which job and when, once per account and job. When 3 different accounts with a confirmed address have reported jobs on the same employer page since the operator last cleared it, every job on it is hidden until the operator has checked it (for a job read from a page, or from the Bundesagentur für Arbeit, that job), and everybody who added or applied to one of its jobs gets a notice, without an email. Employers on the operator's own list are never hidden by reports, only by the operator.
What other accounts are shown. No other account is shown who added a job or reported one: not on the job's page, not after its employer page is approved, and not in any number, since a job only its adders can see is left out of every count shown to anybody else. Six things can be learned without a name. If a job you added from a page is published from somebody else's copy, your notice says so, and links the published job: you learn that somebody else added the same page, though never who. Somebody who added a job from an employer page after you, and whose allowance did not grow when that employer page was approved (the limit below), can sometimes tell that another account added from it first, though never which. If you paste the link of a page that somebody else added and that was then removed on request or found not genuine, you are told so and the page is not read: you learn that somebody else added it, though never who. When pages on one site have been found not genuine more than once lately, the add page refuses that site's links for a while, so you can sometimes tell that other accounts added pages there, though never which. This server also reads pages for every account at one shared pace. When a site asked us to slow down in the last hour, did not answer us in time in the last 15 minutes, or has been asked as much today as we allow, the add page says so; when other pages are being read or are waiting to be saved, it asks you to try again in a minute; and how long it takes to answer can depend on what was read for other accounts before yours. So you can sometimes tell that other accounts pasted links to a site lately, or are adding pages now, though never which. And a job you added from the Bundesagentur für Arbeit that has closed here opens again when the agency's feed lists it again or when another account adds it, so you can sometimes tell that somebody else added it, though never who. The exception is the operator. The review of an employer page lists the handles of the accounts that added jobs there, how many each added and when, and how many of the pages each one first added were approved or found not genuine, because the limit below is about those accounts. The review of a job read from a page names the account that added it, with the same counts, and the accounts that added the same page, with what each saved. Reports reach the operator as a number per job, not as names.
The limit. How many jobs you may add is worked out from the records above: how many you added in the last 24 hours, and how many of the employer pages you were the first to add were approved, or found not genuine or taken down. Once one of yours is found not genuine or taken down, your allowance goes back to the starting one and stays there, and after a set number of them adding links is switched off for your account. Nothing else about the account changes. A job you added from a page counts as an employer page you were the first to add, except that its approval does not raise your allowance. Adding jobs from pages has limits of its own, worked out from the records above too: how many pages were read for your account today, how many of your jobs read from pages are waiting for a check, how many of your links were refused today for where they pointed, and how many of your jobs read from pages were recently declined because they could not be verified. Past one of those, the add page asks you to wait, or adding from pages pauses for the rest of that day or for some days, and the add page says until when.
Websites we count. When a link is refused, or this server reads a page for one, the name of its website is usually counted: the name, how many times, the last day it happened, and, for a page it read, what came of it. No account, no time of day and no other part of the link is kept with it.
People named in a job ad
A job ad can name people: a recruiter, a contact person, or, when the employer is a household or a person, the employer. This section is for them, whether or not they use this service.
The ads read from application systems and public feeds are shown to members as their employer published them, and are kept until the operator prunes the collection (under How long).
When a member adds a job from a web page, what is held about a person the ad names is what the ad says, as the member chose it from the page: the ad's text and the employer's name. For the person who checks the job, and for nobody else, the service also keeps the first 200 characters of each paragraph the member left out, and the titles, employer names and places the page offered; these are kept as the page wrote them, contact details included. Its source is the public web page whose link the member pasted. The lawful basis is a legitimate interest: helping members find and judge job ads, which is what the ads were published for. Email addresses and phone numbers that our scan finds in the job's title, the employer's name and the description are removed before other members can see the job. So are those in the other details read from the page: its places, its kind of employment and its salary's currency. If the operator marks the employer as a private person, when approving the job or when removing it on request, the job's own page shows a private employer in place of the name to everyone but the members who added it, and the name is replaced for good when the job's text is deleted; until then, lists of jobs still show it. The text, with those openings and what else the page offered, is deleted when the job closes here, at the latest thirty days after it was added, or, if it is found not genuine or taken down, 30 days after that. The title, the employer's name (unless it was replaced) and the page's address stay with the job while it is kept, and are deleted with it.
To object, or to have a paragraph or the whole job removed, write to the address on the support page and say which job. On the day we read it, we take that paragraph out of the job's description, or remove the job and delete its text at once; when the employer is a private person, we replace their name at once too. A job already found not genuine or taken down is hidden from everybody already, and a request changes nothing about it: its text is kept the 30 days there are to contest that, and its employer's name is left as it is. The opening of a paragraph the member left out goes only with the whole text, or with a paragraph taken out that holds the same words, so to have another deleted sooner, ask for the job to be removed. We do either without warning anybody and without counting it against the member who added it.
Cookies
One, and it is the sign-in cookie. It is what keeps you signed in and it is not shared with anyone. There is no tracking cookie, so there is nothing to consent to and no banner.
How long
Almost everything is kept until you delete your account, and nothing expires on its own. Said plainly because the opposite is the usual claim: there is no timer that ages out your CV, your preferences, your judgements or your application history. Deleting the account is what removes them.
- Sign-in sessions — at most 30 days, and sooner if you do not come back for 14. Expired rows are checked on every request and swept hourly.
- Sign-up and password-reset attempt records — 1 hour. Every new attempt deletes the expired ones before counting, so the table cannot outlive its own window.
- Add-a-link attempt records — 1 hour. An older one is deleted by the next attempt anybody makes, and all of yours go with your account.
- Page-read and refused-link records — 24 hours, by the same rule: an older one is deleted by the next attempt anybody makes, and all of yours go with your account.
- The jobs you added, and your reports — the record of each is kept until you delete your account, unless the job itself is pruned first (employers' job ads, below), which takes your record of adding or reporting it with it.
- Notices — kept until you delete your account. Dismissing one records when you did; it does not delete it. If the job a notice points at is pruned, the notice stays without it.
- The operator's reviews — kept together with what the checks found. A review of an employer page is kept with no end date: it is about the employer page, it names nobody but the operator who decided it, and it is not deleted with anybody's account. A review of a job read from a page is about that one job, whose employer can be a private person. If the operator marks the employer so, the name is replaced when the job's text is deleted. The words of the ad that the checks quoted are deleted with that text, whoever the employer is. Such a review, like the review of a job from the Bundesagentur für Arbeit, is deleted when its job is pruned (under Employers' job ads), or with your account if the job was still yours alone (under What you can do).
- Your uploaded CV file — deleted when you delete your account, along with everything read out of it. That part is implemented and tested.
- A superseded CV's uploaded file — intended to be deleted some months after you replace it. Still not implemented: replacing your CV supersedes the old version's rows, and the old file stays on the server until you delete your account. Nothing ages it out on a timer, and this notice will not describe a deletion that does not run.
- A CV you change or build on this site — the draft you work on is kept until you publish it, discard it or delete your account; discarding deletes it at once. Once published, it and every version of your CV before it, uploaded, changed or built here, are kept until you delete your account: replacing a CV does not delete the one it replaced. Publishing also sets aside a CV you uploaded that has been read and is waiting for you to check it, which is then kept like any replaced version and is never used to match you.
- Emails you send to the support address — kept in the mailbox at Zoho until the operator deletes them. Nothing in this service deletes them, on a timer or with your account, because the mailbox is outside this service's software; ask, at the same address, to have them deleted. Zoho's own timers then apply. A deleted message goes to the mailbox's trash, which Zoho clears of mail received more than a set number of days earlier (30 unless the mailbox is set otherwise; Zoho allows 5 to 180); Zoho then keeps it on its servers for 30 days more, and in its encrypted backups for three months. The same timer clears the spam folder, so a message Zoho files as spam is deleted without anybody deleting it, possibly before anyone has read it.
- Error reports — 30 days at Sentry, or 90 for one that arrived during a trial (above). Nothing in a report names an account or anything of one, so deleting your account finds nothing in them to delete. Only a report's time, set against this service's own records of who was using it then, could narrow down whose request it was; to have the reports from a time you name deleted sooner, ask at the support address.
- Backups — the fourteen newest nightly copies. Each night a new copy is uploaded, the oldest one past fourteen is deleted from Backblaze, so anything deleted is still in the encrypted backups until fourteen newer copies exist, a deleted account included: a fortnight when every night's backup succeeds, longer when some fail, because it is copies that are counted and not days. Nothing is ever edited inside a backup — a backup you can reach into and alter is not one.
- Jobs read from a page — the copied text (the description, the other choices the page offered, the opening words of the paragraphs left out, and every copy the service made from them) is deleted when the job closes here, at the latest thirty days after it was added; at once when it is declined or removed on request; and 30 days after it is found not genuine or taken down, the time there is to contest that, even when its removal is asked for. What stays is the job without its text: its title, its employer's name (or a private employer), its places, its dates and the page's address, kept as the other jobs you added are (above). When the job is pruned (below), its employer's name and its review go with it.
- Employers' job ads — these are not personal data about you, though your judgements on them are (people an ad names are under their own heading, above). Old ads are removed from the shared collection when the operator prunes it, at a multiple of the freshness window; it is a command someone runs, not a timer, and this notice will not pretend otherwise. Removing an ad removes your judgements on it with it — except where you applied, because your own application record points at it. A job you added by link is pruned the same way, with two more exceptions, because the limit on adding links is counted from them: a job on an employer page found not genuine is kept, hidden, while an account that added it remains, and the job that first brought a new employer page here is kept while the account that added it remains.
What you can do
You can download what this service's database holds about you as one
file, at any time, without asking. That includes the jobs you added by link (submissions.json: when you
added each, its title, the employer, the job board, the link, and whether it is public,
private to the people who added it, or taken down, in which case without the link; for a job
read from a page, also when the page was read, when the job closes here, when its text is
deleted, every field you typed, with what you typed, which fields you chose from the page,
how the page was read (each thing listed under Adding a job from any other page,
above), and, if it was declined, kept private or removed on request, which and why; nothing
in it says whether anybody else added the same job), your reports (reports.csv)
and your notices (notices.csv). It leaves out your password, the secrets
in your sign-in cookie and in email links (only their hashes are stored), and the add-a-link
attempt records and page-read records, which hold only a time each; and, for a job read from
a page, what the page offered and the opening words of the paragraphs you left out, which are
the page's words, and the numbers kept to recognise the same page when it is pasted again;
the file's README lists each and says why. It cannot include the emails you sent
to the support address, which are in the mailbox at Zoho and not in the database; for a copy
of those, ask at the same address.
You can delete your account from the same page — that removes your profile, your CV and everything read out of it, your preferences, your judgements, your applications, your records of the jobs you added, your reports, your notices, and your add-a-link attempt and page-read records, and it is not reversible once the grace period ends. The encrypted nightly backups still hold what it removed until fourteen newer copies exist, a fortnight if every night's backup succeeds, and the copies that hold it are then deleted (under How long, above). It does not reach the support mailbox: emails you sent there stay until the operator deletes them, and at Zoho for a while after that (above), so ask at the same address if you want them gone too.
What happens to a job you added depends on who else can see it. One that only you added and that is still waiting for the operator's check is deleted with your account, unless another account has applied to it or acted on it: then it is taken down for everybody and kept only as that account's record. One that is public, or that somebody else added too, stays, and so does one already taken down; nothing links any of them to you afterwards. If no other account still has a job it added from the same unchecked employer page, the open review of that page is withdrawn, unless reports have hidden the page. Your reports go, but a page they helped hide stays hidden until the operator has decided on it.
A job you added from a page, and that is still yours alone (waiting for its check, or kept private), is deleted with your account, together with the employer's name you saved for it and its review. One a person approved stays public, and the words you typed in it stay with that job, without your name: a typed description until the job's text is deleted, a typed title or employer for as long as the job is kept. One that was declined, found not genuine, taken down or removed on request stays hidden until the collection is pruned, which deletes it with the employer's name saved for it and its review. Until then it keeps what you saved for it and when, but not which account saved it.
You also have the rights the GDPR gives you against any controller: to see what is held, to correct it, to have it erased, to restrict or object to processing, to take it elsewhere, and to complain to a supervisory authority — in Germany, the authority for the state the operator lives in.
Changes
A change to this notice is a new version, and it carries the same version as the terms. When one is published, the next page of your account you open asks you to accept the pair, and until you do, the rest of your account waits behind that page, the download and the delete button included. You stay signed in, nothing is deleted, no email is sent about it, and the service goes on judging postings for you in the meantime. What you accepted is recorded against your account with the date.