On 11 November 2026 the EDPB announced that automated decision-making and AI in recruitment will be the topic of the coordinated enforcement action starting in 2027. Employers that let a tool filter out applicants without a human assessing them take a decision under GDPR Art. 22, which requires an exception, human intervention and an explanation of the logic to the applicant.
At its plenary meeting on 10 November 2026, announced on 11 November, the European Data Protection Board (EDPB) decided that its sixth coordinated enforcement action will concern automated decision-making, including AI-based decision-making, in recruitment. The action will examine how employers comply with the right not to be subject to decisions based solely on automated processing, and with the related right of access, when they use tools that screen candidates automatically. Supervisory authorities join on a voluntary basis, and the work will begin in 2027. It has not been announced whether the Norwegian Data Protection Authority (Datatilsynet) will take part. This article explains when automated selection of applicants is a decision under GDPR Art. 22, which exceptions apply, what the applicant is entitled to know, and what the employer should require from the supplier of the tool.
What will the EDPB examine in its enforcement action on AI in recruitment?
A coordinated action is a joint investigation in which the participating supervisory authorities put the same questions to organisations in their own countries and the EDPB compiles the results in a report. According to the EDPB announcement, the action concerns tools for automated candidate screening, including tools driven by artificial intelligence. The Swedish authority IMY describes the topic in its news item of 16 November 2026 as automated decision-making in recruitment and states that the work will commence in 2027. The investigation may lead the supervisory authorities to open formal cases against individual employers, and such cases may result in orders and administrative fines under GDPR Art. 58 and GDPR Art. 83.
The action reaches further than products marketed as AI. Applicant tracking systems that rank applicants by keywords, exclude applications that do not meet formal requirements, or give each candidate a score after a video interview or a test are all tools for automated screening. Organisations that use recruitment modules in an HR system delivered as SaaS should therefore expect to be covered, even if they have not themselves chosen an AI solution.
When is automated selection of applicants a decision under GDPR Art. 22?
Under GDPR Art. 22(1), the data subject has the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her. The rule applies in Norway through Personal Data Act § 1. Recital 71 of the General Data Protection Regulation expressly mentions e-recruiting practices without any human intervention as an example of such a decision. In our assessment, the rejection of a job application meets the requirement of significant effect. The basic rules on the processing of personal data are collected on our data protection topic page.
The decisive condition in practice is whether the decision is based solely on automated processing. In C-634/21 SCHUFA, the Court of Justice held that an automatically calculated probability value may itself be a decision where the person taking the final decision draws strongly on that value. Applied to recruitment, this means that a score or ranking produced by the tool may be the decision itself, even though a recruiter formally takes the decision.
The most common risk lies in how the tool is used. An employer that reads only the 20 highest-ranked applications has in reality let the tool determine the outcome for all the others. For those applicants there is, in our assessment, a decision based solely on automated processing, regardless of how thoroughly the first 20 are assessed. Human involvement counts only where the person assessing has the competence, access to the underlying information and authority to reach a different result. We have examined that requirement in more detail in our article on automated decisions after the Uber fine.
Which exceptions make automated screening lawful?
Automated decisions are permitted only in the three cases listed in GDPR Art. 22(2). The decision must be necessary for entering into or performing a contract, be authorised by Union or Member State law with suitable safeguards, or be based on the data subject’s explicit consent. Norwegian law contains no specific legal basis for automated selection of applicants by private employers.
Consent is a weak basis in an application process. The applicant is in a position of dependence on the employer and will seldom be able to refuse without fearing that the application is weakened, and consent is then not freely given. The contract exception is therefore the practical basis. The Article 29 Working Party’s guidelines on automated decision-making (WP251rev.01), which the EDPB has endorsed, take the view that necessity may exist where a position attracts so many applications that it is not practically possible to identify suitable candidates without automated means. The employer must be able to document that less intrusive methods would not suffice.
Where the decision rests on the contract exception or on consent, the employer must implement suitable measures under GDPR Art. 22(3). The applicant must at least have the right to obtain human intervention, to express his or her point of view and to contest the decision. Under GDPR Art. 22(4), the decision may also not be based on special categories of personal data, such as health data or data revealing racial or ethnic origin, unless the narrow conditions in that provision are met.
The Commission’s Digital Omnibus proposal would recast Art. 22 from a right into a list of cases in which automated decisions are permitted. The Council’s negotiating mandate of 21 October 2026 reinstates the wording as a right and retains the requirements of human intervention and the right to contest the decision. At the same time, the mandate keeps the Commission’s addition to the contract exception, under which a decision may be necessary regardless of whether it could be taken otherwise than by solely automated means. If adopted, this would reduce the weight of the requirement to document that less intrusive methods would not suffice. The amendments have not been adopted, and the current Art. 22 is the basis for the enforcement action. The process is covered in our article on the Digital Omnibus and the GDPR.
What can the applicant require to know about the logic of the tool?
At the time the data are collected, the applicant must be told that automated decisions are taken and be given meaningful information about the logic involved and the envisaged consequences, pursuant to GDPR Art. 13(2)(f). The same information may be requested through access under GDPR Art. 15(1)(h), which is precisely the right of access the EDPB has included in the enforcement action. A general sentence in the privacy notice stating that “AI may be used in recruitment” does not meet the requirement.
The Court of Justice clarified the content of the right in C-203/22 Dun & Bradstreet Austria. The data subject is entitled to a concise and intelligible explanation of the procedure and principles actually applied, so that he or she can understand which data were relied on and how they affected the result. Trade secrets are not a general ground for refusing access. The Court held that a national rule which as a rule excludes access in order to protect trade secrets is contrary to the Regulation, and that the balancing must be carried out case by case. Deadlines and exceptions for access requests are reviewed in our article on subject access requests under the GDPR.
The obligation falls on the employer as controller, but it is the supplier that knows the model. An employer that cannot itself explain which criteria the tool weighs will not be able to answer an access request within the one-month time limit. This must therefore be resolved in the contract with the supplier before the tool is put into use.
How does Norwegian anti-discrimination law affect the use of AI in recruitment?
The prohibition of discrimination in Equality and Anti-Discrimination Act § 6 applies to all aspects of an employment relationship, including advertising and appointment, pursuant to Equality and Anti-Discrimination Act § 29. Under Equality and Anti-Discrimination Act § 30, the employer may not, during the recruitment process, obtain information about matters including pregnancy, religion, ethnicity and disability. A tool that infers such matters indirectly, for example from names, place of residence, gaps in the CV or a recorded video interview, may lead to a breach even though the employer never asked for the information.
The rule on the shared burden of proof in Equality and Anti-Discrimination Act § 37 turns this into a concrete litigation risk. Where there are circumstances giving reason to believe that discrimination has occurred, discrimination is deemed to have occurred unless the employer shows otherwise on the balance of probabilities. An employer that cannot show how the tool assessed the applicants will have difficulty rebutting that presumption. In addition, an applicant who believes he or she has been passed over may require written information about the qualifications of the person appointed, pursuant to Equality and Anti-Discrimination Act § 31.
What does the AI Act say about recruitment tools?
In Annex III point 4(a), the AI Act classifies AI systems intended to be used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, to analyse and filter job applications and to evaluate candidates, as high-risk systems. The employer is the deployer and must, among other things, assign human oversight to persons with the necessary competence and authority, pursuant to AI Act Art. 26(2). Under AI Act Art. 26(11), deployers of such systems must inform the persons concerned by the decisions that the system is being used. Under AI Act Art. 26(9), the information from the provider must be used in the data protection impact assessment.
Following the amendment made by Regulation (EU) 2026/1744, the obligations for high-risk systems in Annex III apply in the EU from 2 December 2027. The AI Act has not yet been incorporated into the EEA Agreement, and the obligations therefore do not apply in Norway today. The 2027 enforcement action is based on the GDPR, which already applies in full. How AI vendor terms for high-risk use allocate responsibility is discussed in our article on AI vendor terms for high-risk use.
What should the employer require from the supplier of the screening tool?
The supplier of a recruitment tool is normally a processor for the employer, and the contract must comply with GDPR Art. 28. Beyond a standard data processing agreement, the employer should make sure that the contract provides the basis needed to meet its obligations towards applicants. The table sets out the main requirements.
| Requirement for the supplier | Reason |
|---|---|
| Written description of the criteria and weighting used by the tool | Information and access under GDPR Arts. 13 and 15, see C-203/22 |
| A function allowing a recruiter to view, override and reverse each assessment | Human intervention under GDPR Art. 22(3) |
| A log of the data underlying each assessment | Responding to access requests and evidence in a discrimination complaint |
| Documentation of testing for bias between groups | The burden of proof in Equality and Anti-Discrimination Act § 37 |
| Assistance with the data protection impact assessment | GDPR Art. 35, and later AI Act Art. 26(9) |
| A prohibition on using applicants’ data to train the supplier’s models without a specific agreement | Purpose limitation and the supplier’s role as processor |
The requirements for documentation of the logic and for bias testing are those most often missing from standard terms. The supplier will often invoke trade secrets. Since the Court of Justice has held that trade secrets do not provide a general exemption from the right of access, the contract should at least oblige the supplier to provide an explanation that the employer can pass on, and to make full documentation available to the supervisory authority. Terms for data processing in SaaS are reviewed in our article on data processing agreements for SaaS.
What should the organisation do before the enforcement action begins?
Employers that use automated screening have some months before the action begins, and most of the work concerns documentation. The measures below are based on what the EDPB has said the action will examine.
- Map every step of the recruitment process where a system sorts, ranks, scores or excludes applicants, including functions in HR systems that are not marketed as AI.
- Determine for each step whether a person with authority actually assesses all applicants, or whether the tool in practice determines the outcome for some of them.
- Document the exception under GDPR Art. 22(2) where the tool decides, and establish a channel through which the applicant can request a human assessment and contest the result.
- Carry out or update the data protection impact assessment. Datatilsynet’s list of processing operations requiring a DPIA covers processing whose purpose is to offer a service for commercial use that involves predicting, among other things, job performance, and the method is described in our article on DPIAs for AI tools.
- Update the information in job advertisements and the application portal so that it explains which criteria the tool weighs and what significance the assessment has for the applicant.
- Renegotiate the supplier contract in line with the table above before the next renewal, and require the supplier to deliver, within a fixed deadline, the explanation needed to answer an access request under GDPR Art. 15(1)(h).
Questions and answers
Is it sufficient for a recruiter to approve the shortlist proposed by the tool?
No, not where the approval is a formality. The person assessing must have the competence, access to the underlying information and authority to reach a different result. Applicants who were sorted out by the tool and whom nobody reviews have in reality received a decision based solely on automated processing.
Can the employer base automated selection on the applicant's consent?
That is rarely workable. Consent must be explicit and freely given, and a job applicant will seldom be able to refuse without weakening the application. The exception for processing necessary for entering into a contract is normally the relevant basis, and the necessity must be documented.
Does the AI Act apply to recruitment tools in Norway today?
No. The AI Act has not yet been incorporated into the EEA Agreement. In the EU, the obligations for high-risk systems in Annex III, including recruitment tools, apply from 2 December 2027. The GDPR and the Norwegian Equality and Anti-Discrimination Act already apply in full.
- EDPB, EDPB adopts statement on review of the Republic of Korea adequacy, processor BCR recommendations, and selects topic for 2027 coordinated enforcement action (11 November 2026)
- IMY, Automatiserat beslutsfattande i fokus för EDPB:s kommande samordnade åtgärd (16 November 2026)
- General Data Protection Regulation (EU) 2016/679 Arts. 12, 13, 15, 22, 28, 35, 58 and 83, recital 71
- Norwegian Personal Data Act (personopplysningsloven) § 1
- Norwegian Equality and Anti-Discrimination Act (likestillings- og diskrimineringsloven) §§ 6, 29, 30, 31 and 37
- AI Act (EU) 2024/1689 Art. 26 and Annex III point 4
- Regulation (EU) 2026/1744 (digital omnibus on AI)
- Judgment of the Court of Justice of 7 December 2023, C-634/21 SCHUFA Holding (Scoring)
- Judgment of the Court of Justice of 27 February 2025, C-203/22 Dun & Bradstreet Austria
- Article 29 Working Party, Guidelines on Automated individual decision-making and Profiling (WP251rev.01)
- Datatilsynet, Når må man gjennomføre en vurdering av personvernkonsekvenser?
- Council of the EU, Digital rulebook: Council agrees position to simplify data protection, privacy and cyber reporting rules (21 October 2026)
Next legal review: 1 February 2027