HR clicking 'Reject' Is Not a Human Decision

Europe's data protection authorities will examine AI in recruitment in 2027. What your company must show when software ranks and rejects candidates.

PublishedReading time: 7 min read

A scientist interacts with a robot helper, demonstrating modern technological innovation.

Europe's data protection authorities have announced that in 2027 they will jointly examine automated decision-making in recruitment, including AI-driven CV screening. If your applicant tracking system ranks candidates and sends rejections on its own, that review concerns you. Here is where an assistant ends and a decision-maker begins, and what to have ready before the first questionnaire lands.

What the European Data Protection Board Announced

On 11 November 2026, the European Data Protection Board (EDPB) selected the topic for its sixth coordinated enforcement action: automated decision-making in recruitment, including AI-driven decisions. Authorities will focus on the right not to be subject to a decision based solely on automated processing, and on the related right of access, where companies use automated candidate screening tools. National authorities will sign up voluntarily in the coming weeks, and the action itself launches during 2027.

Earlier rounds followed a simple pattern: authorities contacted selected controllers, asked how they had set up the obligation in question and published the results in a joint report. In 2024 the topic was the right of access; in 2025, the right to erasure. This time it is hiring, an area where software makes more of the call than most boards realise.

On 8 October 2026, the Dutch data protection authority fined Uber EUR 824,990,000 because its systems automatically deactivated drivers' accounts and drivers were not properly informed. Uber has appealed and the decision is not final. The logic of the case maps directly onto recruitment: an algorithm decided whether a person could earn a living, and nobody else stepped in.

Where the Assistant Ends and the Decision-Maker Begins

Take an ordinary scenario. A company with 300 employees advertises a sales role and receives 400 applications. The applicant tracking system gives each candidate a match score. The recruiter opens the top thirty profiles; the remaining 370 people receive an automated email thanking them and turning them down. For 370 people, the algorithm settled the outcome.

Article 22(1) of the General Data Protection Regulation (GDPR) gives individuals the right not to be subject to a decision based solely on automated processing that produces legal effects concerning them or similarly significantly affects them. Recital 71 names e-recruiting practices without any human intervention as an example. A rejected application is therefore a significant decision.

The usual defence is: "A person decides here; the system only recommends." The Article 29 Working Party guidelines on automated decision-making, endorsed by the EDPB, require human involvement to be meaningful rather than a token gesture. It must come from someone with the authority and competence to change the decision. If a recruiter confirms fifty rejections with one click without opening a single profile, that is a rubber stamp, not a decision.

Pointing to the vendor does not help either. In SCHUFA (C-634/21, 7 December 2023), the Court of Justice of the EU held that a score produced by a third party can itself be a decision under Article 22 where the outcome depends decisively on it. The controller of your candidates' data is your organisation, not the software provider.

What Regulators Will Want to See

Article 22(2) GDPR permits solely automated decisions in only three cases: where necessary for entering into a contract, where authorised by law, or with explicit consent. Candidate consent is a weak footing, given the imbalance between employer and applicant. Contractual necessity has to be explained: why human review is not feasible at the volume involved. Even then, Article 22(3) requires you to let candidates obtain human intervention, express their point of view and contest the decision.

Information up front

Under Article 13(2)(f), candidates must be told when their data is collected that automated decision-making takes place, and be given meaningful information about the logic involved and the consequences for them. A generic sentence in the privacy notice on your careers page will not do.

Answering an access request

A rejected candidate has the right under Article 15(1)(h) to find out how the system decided about them. In Dun & Bradstreet Austria (C-203/22, 27 February 2025), the Court of Justice clarified that the controller must explain, in an intelligible way, the procedure and principles actually applied to that person's data. Citing the vendor's trade secrets does not automatically rule out an answer. The deadline is one month, and a candidate's request belongs in your GDPR request queue like any other.

An impact assessment

Systematic evaluation of people based on automated processing that leads to significant decisions is a textbook case under Article 35(3)(a) GDPR where a data protection impact assessment (DPIA) is mandatory. If you have none for your recruitment tool, expect it to be one of the first questions.

The AI Act Adds a Second Layer

The AI Act (Regulation (EU) 2024/1689) classifies AI systems used for recruitment or selection, in particular to target job adverts, analyse and filter applications and evaluate candidates, as high-risk (Annex III, point 4). Following the AI Omnibus, the rules for these systems apply from 2 December 2027.

A company using such a tool is a deployer. Article 26 will require it, among other things, to use the system in line with the instructions, assign human oversight to people with the necessary competence, training and authority, inform workers' representatives and affected workers in advance, and inform the people the system decides about or helps decide about. Article 86 gives affected persons a right to an explanation of individual decisions.

One rule already applies today. Emotion recognition in the workplace has been prohibited under Article 5(1)(f) since 2 February 2025, and the Commission's guidelines on prohibited practices extend the workplace to the recruitment process. If your video interview tool scores a candidate's "enthusiasm" or "nervousness", switch that feature off now.

In Czechia, the GDPR is supervised by the Office for Personal Data Protection (ÚOOÚ), and the Employment Act also limits what employers may ask applicants. A tool that infers from a CV what you may not ask can cause problems beyond the GDPR.

What This Means for the Board

Recruitment is often left to HR and software procurement. Deciding who gets through the door, however, is the controller's responsibility, which sits with the company as a whole. If leadership cannot say, when asked directly, which tools decide about candidates and who checks their output, it will struggle to show it acted with due care. The good news: most of the fix is a process change, not a system replacement.

Checklist Before 2027

  • Map your tools. ATS, job board filters, online assessments, video interviews. For each, record whether it scores, ranks or rejects.
  • Find the points with no human. Where can a candidate be rejected without anyone opening their profile?
  • Set up real review. Who decides, with what authority, what they see and how they record the reason. That record is your evidence.
  • Check your legal basis. If you keep automation, document why it is necessary and put the Article 22(3) safeguards in place.
  • Rewrite candidate information. Be specific: what the system assesses and what that means for the applicant.
  • Prepare your access request answer. Have a template explaining the logic, and the vendor's input, before the first request arrives.
  • Complete a DPIA. For the recruitment tool, not for HR in general.
  • Switch off emotion recognition. The ban already applies.
  • Train your recruiters. Regulators will not accept human oversight without training.
  • Review the vendor contract. Documentation, instructions for use, logs and support with candidate requests.

Where VPoint Helps

This work sits between HR, GDPR and AI governance, so it often gets lost across three tools and three people. VPoint keeps it in one place. Candidate access requests run in a queue with the statutory deadline and an audit trail, with your processing records and the DPIA for the recruitment tool alongside. The platform uses AI to draft recruiter training straight from your own recruitment policy and records a completion date for each person. Our AI governance module, with an AI system inventory and Annex III classification, is in development, and a readiness review is available now.

Book a 30-minute demo and we will show you how to get your hiring process ready for the 2027 review. One annual price per organisation, with no per-user fees.