Trust centre

AI you can explain. Decisions you own.

Recruitment is a high-risk use of AI under the EU AI Act. We built Doowings Hire for that from the first screen: the safety is in the interface, where you and your auditors can see it.

EU AI Act

Every obligation, mapped to a feature.

The Act asks providers and users of high-risk AI for oversight, logging, transparency and good data. Here is where each one lives in the product.

ObligationWhat it asksWhere it lives in Doowings Hire
Human oversightPeople must be able to oversee, override and stop the AI.Doowie never sends a message or moves a candidate. Drafts wait in a review queue; agent actions are cards you Apply; dismissals and stage moves are human clicks with undo.
Record-keepingThe system must log events so its use can be traced.Every AI call and every search is logged — purpose, who asked, who confirmed — without raw personal data. AI-call logs are kept two years, search logs one year.
Transparency to usersUsers must understand the output and use it appropriately.Matches show criteria, evidence and source; a match is labelled as fit to criteria, never a hiring probability. Anything Doowie wrote carries its mark.
Information to candidatesPeople affected must be told AI is used.The careers site shows a plain-language AI notice before collection. Workspaces configure a transparency contact and notice templates; first-contact emails carry an editable privacy notice.
Data governanceData used must be relevant, lawful and minimised.A lawful basis per person (consent or legitimate interest), blocked and do-not-contact people excluded from AI processing and outreach, configurable retention, and personal data minimised before AI processing.
Robustness & securityThe system must resist errors and manipulation.Text from CVs and documents is treated as untrusted and cannot instruct the AI. Workspaces are isolated at database level; sign-in uses MFA and re-authentication for sensitive changes.

Built for the EU AI Act. We don’t call the product “certified”: no such certificate exists for recruitment AI today, and our own conformity work — technical documentation, risk assessment and bias testing — is in progress. Ask us for its current state.

Human in the loop

Doowie drafts. A person decides. Always.

This isn’t a setting you can forget to turn on. Outreach drafts wait in a review queue; the assistant’s actions are proposals; reply analysis only suggests stopping a sequence. Our test suite checks these rules on every change.

Audit trail

A record your DPO can read.

What the AI did, for which purpose, and who confirmed it — on the candidate’s timeline and in the logs, exportable when a client, a candidate or an auditor asks.

GDPR

GDPR handled in the product, not in a ticket.

Lawful basis, enforced

Consent for applicants, legitimate interest for sourced people — tracked per person. Blocked and do-not-contact people are excluded from AI processing and outreach.

Retention you configure

Windows for applicants, sourced people and the talent pool; a live preview; an “erasing soon” list; extensions only with a written reason, and capped.

Access & portability

A full export of a person’s data from their profile, in one click.

Erasure with a certificate

Complete erasure across records, files and history — with a certificate that itself holds no personal data.

Objection that follows the person

An objection applies across the same person’s candidate and contact records, and a suppression record survives erasure so they aren’t re-imported.

Legitimate-interest assessment

Admins attest a versioned assessment in the product before opt-out talent pooling is enabled.

Where data lives

Your data stays in Europe.

  • Hosted in the EU — the application, the database, CV files, background processing, error monitoring and system email.
  • CVs and search processed in the EU — reading CVs and the search index run on EU-hosted AI.
  • Assistant and drafting — we minimise the personal data sent and log every call. These features are moving to EU-hosted processing; ask us for the current state.
  • Our sub-processors are listed in our data processing agreement.

Security

Secure by default.

Workspace isolation

Every query is scoped to your workspace, and isolation is enforced again at database level.

Strong sign-in

Authenticator-app MFA for every user, re-authentication for sensitive changes, sessions you can sign out everywhere.

Least privilege

Five roles and a visible permission matrix. Deactivating a user disconnects their mailbox and LinkedIn.

AI that can’t be hijacked by a CV

Document text is isolated from instructions, so a CV can’t tell the AI what to do.

Responsible disclosure

Found a vulnerability? Our security page and security.txt explain how to reach us.

Certifications — honest status

We don’t show badges we haven’t earned. No ISO 27001 or SOC 2 report today; we’ll say so here when that changes.

Bring your DPO.

We’ll show the notices, the oversight and the audit trail on real screens, and send our DPA and sub-processor list ahead of the call.