Governance

Answers for the procurement review.

Banks, telecoms and public institutions ask the same questions before an AI goes near candidate data. These are the answers — including the ones that are still a roadmap rather than a fact.

Back to overview

01AI accountability

The AI has authority to recommend, and none to decide.

Automated decisionsNone. The AI ranks, scores and explains. Advancing or declining a candidate is an action by a named user, and it is written to the audit log.

ExplainabilityEvery score carries a link to what it was scored from — a screening answer, a line of the interview transcript, or a verified record. Unsupported criteria are marked unverified rather than filled in.

Human oversightConfigurable per workspace. At the strictest level a reviewer cannot record a decision until they have opened the evidence, and any override of an AI rating needs a written reason.

Audit trailEvery consequential action is logged, the AI included. Automatic pipeline transitions are marked as such, so a machine step is never mistaken for a human one.

Model useCandidate data is processed for the hiring purpose it was collected for. It is not pooled across customers and is not used to train models.

02Data protection

Where the data sits and who can reach it.

Tenant isolationEvery query is scoped to one workspace; there is no cross-tenant read path. The boundary is covered by an automated test that fails the build if it is crossed.

Access controlRole-based permissions down to the action, with separate read and write rights. Deactivating a member ends their sessions immediately.

Single sign-onOIDC with just-in-time provisioning and role mapping from your identity provider.

SecretsIntegration credentials are stored encrypted, and traffic is TLS-only.

RetentionRetention periods are set per workspace; expired candidate data is removed on a schedule rather than left to accumulate.

DeploymentManaged cloud, or a dedicated instance for organisations that require one.

03Candidate rights

The candidate is a party here, not a record.

Consent-gated identity

Employers see an anonymised profile until the candidate agrees to be identified. Every reveal is logged.

Anonymous and paused

A candidate can browse anonymously or pause visibility entirely, at any time.

Export and erasure

Their data can be exported or deleted on request, including assessment answers and transcripts.

Human review

Candidates are told AI is part of the process and can ask for a person to review the outcome.

04Regulation

Written for the rules that are coming, not only the ones already here.

EU AI ActRecruitment is a high-risk use. We hold a documented risk-management file covering intended purpose, known limitations, human oversight and monitoring.

Data protectionProcessing is tied to a stated recruitment purpose and to candidate consent, with the rights above implemented in the product.

TransparencyAI involvement is disclosed to the candidate before the process begins, not in a footnote afterwards.

05Not yet

What we cannot claim today.

A security page that only lists strengths is not useful to a reviewer. These are open items, stated plainly so you can plan around them.

SOC 2 / ISO 27001In progress, not certified. We will share the current control set and the timeline on request.

Two-factor for workspace usersAvailable today for platform operators. For workspace users, two-factor comes from your identity provider through SSO; native TOTP is on the roadmap.

Independent penetration testScheduled. Results will be shared with enterprise customers under NDA.

06Contact

Found something, or need the full control set?

Security questionnaires, the risk-management file and vulnerability reports all reach the same place. We answer within one business day.

See it run on one of your open roles.

A working session, not a slide deck: we take a real vacancy, define it in the product, and walk through the shortlist and the evidence behind every score.

Jobs