EU AI Act · High-risk AI (Annex III, point 4)

How CVsprings approaches the EU AI Act

AI systems used for recruitment and candidate filtering are classified as high-risk under Annex III, point 4 of the EU AI Act. Provider obligations (Articles 9–15) apply from 2 December 2027. CVsprings is building toward full conformity ahead of that date: the features and documents on this page describe what the product does today, and what remains in progress.

What this means in practice

CVsprings is an advisory candidate-fit scoring tool. It produces scores that support a recruiter's screening work; it never makes or executes a hiring decision. Our design approach maps each relevant AI Act obligation to a concrete product behaviour — the table below lists only what is actually implemented.

Formal conformity assessment (including a quality management system, test protocols, and technical documentation per Annex IV) is in progress — items not yet complete are explicitly labelled Roadmap in our documentation rather than claimed.

Feature → obligation mapping

Obligation	What CVsprings does today
Risk management (Art. 9)	A documented risk log with identified risks, mitigations and review dates; an anonymization option that strips contact details and name lines before scoring; a human-decision-only design (no score is ever converted into a decision automatically).
Data governance (Art. 10)	Uploaded CV files are deleted from the server immediately after text extraction — no permanent CV storage. When Anonymize is enabled, contact details, URLs, address lines, early years and standalone name lines are stripped from the text before scoring. No demographic data is collected.
Record-keeping (Art. 12)	An audit log in which scoring data is immutable after creation (only the recruiter's decision and note can change, and every change is recorded in an append-only change history). Each record stores the scores, weights, app version, scoring-engine version, analysis timestamp, and the email of the reviewer who saved it.
Transparency (Art. 13)	Published instructions for use; an in-app “Why this score?” explanation for every analysis; a published scoring methodology; and a candidate notice template (below) that clients can give to applicants.
Human oversight (Art. 14)	No automated decisions anywhere in the product: decisions exist only as an explicit recruiter input. The UI labels every score as a decision aid, and each saved record attributes the decision to the signed-in reviewer.
Accuracy & robustness (Art. 15)	A published scoring methodology (deterministic rules applied identically to every CV in a batch), recruiter-configurable weights that are stored with each record, and a per-organization scoring-consistency monitor on the bias & monitoring page.

What we provide to clients

Alongside the in-product features, CVsprings maintains a compliance document pack for clients and their advisers:

📖Instructions for use
Intended purpose, correct operation, the human-oversight requirement, known limitations, and the deployer's own duties (Art. 13).
⚠️Risk management log
Identified risks, mitigations, residual risks and review dates (Art. 9).
🔧Technical documentation outline
Annex IV working document: architecture, scoring pipeline and logging design, with open items explicitly marked TODO.
📝Candidate notice (EN + NL)
Ready-to-paste text informing applicants that an AI-based tool supports the screening process.
📄Compliance pack cover
One-page summary of the product, its risk classification, the 2027 timeline and the included documents.

Candidate notice template

Clients can paste this into job postings or applicant communications to meet their own transparency duties:

English

As part of our selection process, applications may be assessed with the support of an AI-based tool (CVsprings) that compares CV content against the job requirements. The tool produces an advisory score only; all decisions are made by our recruiters. You may request information about this process or ask for a human-only review by contacting us.

Nederlands

Als onderdeel van onze selectieprocedure kunnen sollicitaties worden beoordeeld met ondersteuning van een AI-tool (CVsprings) die de inhoud van het cv vergelijkt met de functie-eisen. De tool geeft uitsluitend een adviserende score; alle beslissingen worden genomen door onze recruiters. U kunt informatie over dit proces opvragen of verzoeken om een beoordeling door uitsluitend een mens door contact met ons op te nemen.

This page describes CVsprings’ design approach and is not legal advice. Formal conformity assessment is in progress toward the December 2027 deadline.