Data processors and AI provider

Last updated 18 October 2026. This page supplements our privacy policy and describes openly who processes data on our behalf, what they receive and on which terms. We update the page when things change. This is a translation of the Danish page, which applies if the two differ.

1. Overview

Elsepath runs on servers at Hetzner in the EU (Germany and Finland) with encrypted disks. Mails from us are still sent from our own mail server, which is why there is no mail service on the list.

Planned and named here for transparency: payment (Stripe), when payment is activated. If we change AI provider or move the operation to another provider, it happens only when the data processing agreement is in place and this page has been updated.

2. What do we send to the AI provider, and what happens to it?

When you use the AI functions (match assessment, "Create CV & application", the revision chat, the support chat), we send the necessary extract per call over an encrypted connection (TLS): relevant CV content, the job posting's text and your instructions. We never send your password, your e-mail address is not needed for the calculation, and we never send the whole database, only the content of the specific call.

According to Ollama's privacy policy (last updated March 2026, reviewed by us on 28 July 2026 and again on 8 August 2026), the following applies to their cloud models:

• The content is processed temporarily and not stored beyond the time it takes to answer the call. They use technical measures designed to minimise the retention of prompt and response content.
• The provider passes the same requirements on to its hosting partners: no logging, no training and zero data retention. The requirement applies to the cloud service as a whole, not only to selected models.
• Your content is never used to train AI models.
• The metadata they collect (for example app version and number of calls) does not contain the content of prompts or responses.
• They do not sell personal data.
• In case of a security breach they notify the supervisory authority within 72 hours, unless the breach is unlikely to result in a risk.
• The policy recognises GDPR rights for persons in the EU, including the right to lodge a complaint with a supervisory authority.

Sources at the provider: Ollama's privacy policy and terms. We encourage you to read them yourself; they are the provider's own documents and may change.

3. Status: processing in the USA and basis of agreement

Ollama processes cloud calls in the USA and today offers neither an EU region nor a public data processing agreement under Article 28 of the GDPR, and at our latest check (28 July 2026) we did not find them on the EU-US Data Privacy Framework list. This means that the formal basis of agreement for this part does not yet meet the level we want.

In August 2026 the provider announced serving capacity in both the USA and Europe. We are looking into whether our calls can be run in Europe; until that is confirmed, this page describes the processing as based in the USA.

If you do not want your content sent to the AI provider, you can use Elsepath without the AI functions: you can see the jobs found without a match score and refrain from using "Create CV & application" and the chat.

4. Backup at Google

For operational security, daily backup copies of the database are copied to Google Drive under Google's data processing agreement for Workspace. The whole file is encrypted with our public key before it leaves the server, and CVs, applications, posting texts and mail passwords are also field encrypted (AES-256-GCM) inside it. The encryption keys are kept separate from the backups. At Google, daily copies are kept for 14 days and Sunday copies for 8 weeks. Backups are used only to restore after a breakdown, never to look up individuals' data.

5. Job sources and our crawler (ElsepathBot)

The job postings in the service come from public sources: Jobnet (public API), Jobindex (official RSS feeds), open ATS APIs (for example Greenhouse) and Danish companies' own, publicly available career pages. These are company and job data, not personal data about our users.

The career pages are visited by our own crawler with the identifiable User-Agent ElsepathBot, which links to this page. It follows these rules:

• It respects robots.txt on each domain and never fetches pages excluded for crawlers. Domains with a total ban are not visited at all.
• It fetches only public pages: never content behind a login and never LinkedIn.
• It runs with a low load: one call at a time to the same domain, at least three seconds between calls, and each site is visited at most once a day.
• If a site answers "access denied" (HTTP 403 or 401), or robots.txt forbids us the job pages, we leave the site alone and try again only a month later. We never try to get around a no.
• Contact persons may appear in a posting (name and e-mail in the job advert). They are processed only as part of the published posting and shown only in the posting's context. The posting is deleted from our shared job pool 30 days after it has expired; if a user has the job on their list, it stays there until the user deletes their account. See the privacy policy, section 10.

If you are responsible for a career page and want it excluded, you can either update your robots.txt (we follow it within a day) or write to [email protected].

6. What we never do

• We never sell your data and show no advertising.
• We never send applications or anything else on your behalf; you always send yourself.
• We do not share your data with employers, headhunters or other users.
• We never take a new data processor into use without updating this page and the privacy policy.

Questions? Write to [email protected].