You receive an automated rejection minutes after sending your CV. Did a recruiter read your application, or did software filter it using keywords and scores? The answer matters, particularly as artificial-intelligence tools become a routine part of recruitment.
The 2026 debate has an important distinction: some candidate rights already exist under the GDPR, while AI Act obligations specific to high-risk recruitment systems will apply at a later date.
What an algorithm can do in recruitment
Software can rank CVs, extract work experience, compare a role’s requirements with a candidate’s profile, or suggest interview questions. Some systems use fixed rules; others use statistical models or generative AI. They do not all have the same degree of complexity, and not every automated form is legally an artificial-intelligence system.
For more on artificial intelligence, see “AI-resistant” jobs: what matters more than a profession’s name, according to the data, which provides further context on the subject.
The issue arises when assessment becomes opaque. A candidate may be screened out because a term does not appear exactly in their CV, because their employment history includes a gap, or because the model has learned correlations that disadvantage certain people. A statistical correlation is not automatically a fair selection criterion.
Rights you already have in 2026
The General Data Protection Regulation, or GDPR, also applies to recruitment. An employer must have a legal basis for processing personal data and must inform candidates in accordance with the applicable obligations.
Article 22 of the GDPR addresses decisions based solely on automated processing that produce legal effects or significantly affect an individual. The framework is nuanced, with exceptions and safeguards. Not every CV rejected with the help of software automatically falls into this category, especially where there is genuine human involvement.
Where processing is disputed, a candidate may request the information to which they are entitled, seek correction of inaccurate data and, subject to the law, lodge a complaint with the data-protection authority. The request should be framed precisely, without assuming that every employer is required to publish an algorithm’s full code.
When the AI Act comes into play
In material updated on 6 October 2026, the European Commission clarified the AI Act’s phased implementation timetable. From 2 August 2026, certain provisions became applicable, including prohibitions on AI practices and transparency rules for defined categories of systems.
For high-risk systems in Annex III—a category that includes certain uses of AI in recruitment and selection—the date indicated by the Commission is 2 December 2027. It would therefore be incorrect to say that all requirements specific to this category became mandatory in August 2026.
This distinction does not give employers a “year without rules”: employment law, data-protection rules and the ban on discrimination continue to apply.
What real transparency looks like
A useful explanation is not the statement, “we use AI to improve the candidate experience”. Candidates need to know whether their data are being analysed automatically, what types of information are relevant, and which channel is available for correcting errors or challenging a decision where the law provides that right.
For employers, practical transparency means taking stock of the tools in use, assessing data quality, testing for discrimination risks and defining human oversight that is more than a formal sign-off.
Five things to check when applying
Read the recruitment platform’s privacy notice. Use a CV with clear job titles, dates and demonstrable skills, without adding unnecessary personal information. Check whether the platform allows an application to be corrected. Keep confirmation of your application. If you suspect a solely automated decision with a significant effect, ask for clarification in terms of the rights available to you.
There is no value in filling your CV with keywords unrelated to your experience. An application that software can read must still be persuasive to a human being.
The next test is about trust, not technology
In recruitment, the question is not only how quickly an algorithm finds the right candidate. It is also whether a rejected candidate can understand the process well enough to identify an error or an unfair outcome. Automation can help with selection, but it should not be mistaken for objectivity.
Research updated on 14 October 2026.











Discuția cititorilor
Ai o opinie sau o întrebare despre acest articol? Spune-ne ce crezi.
Fii primul care deschide discuția.
Te rugăm să păstrezi un limbaj civilizat și să discuți la subiect. Comentariile spam sunt filtrate automat.