Algorithm Denied Your Job, Loan, or Benefit? You Can Demand an Explanation

A refusal of employment, credit, or a benefit increasingly happens without any human involved, an algorithm evaluates the application according to its own criteria. The EU AI Act introduces a concrete tool for such situations: Article 86 grants anyone affected by such a decision the right to demand a clear explanation of what role the AI system played and which factors determined the outcome.

What Article 86 Provides

The provision applies when a decision based on the output of a high-risk AI system produces legal effects or similarly significantly affects a person's situation, and that person considers the decision unfavorable to them. In such a case, they can demand that the entity deploying the system provide a clear and understandable explanation of the algorithm's role and the main elements of the decision made.

The obligation does not require a company or institution to disclose the model's full code or the technical details of how a neural network operates. It is enough to provide information that allows the person to understand which data and circumstances determined the outcome. A response along the lines of 'refused for failing to meet internal criteria' may prove insufficient if the person concerned does not learn which data or circumstances were decisive and how they might improve their situation.

Recruitment and Employment

The AI Act classifies AI systems used in employment, worker management, and access to self-employment as high-risk systems. This applies particularly to algorithms that filter job applications, assess performance, allocate tasks, or support decisions to terminate employment.

A job candidate whose application was rejected by a CV-scanning algorithm can demand that the prospective employer explain which parameters and criteria led to the rejection. Similarly, an employee subject to an unfavorable HR decision based on the output of an AI system can request an explanation.

Loans and Insurance

Credit scoring has been added to the EU's list of high-risk systems, as have systems assessing risk and setting premiums in life and health insurance. In Poland, banks have already been required to explain loan refusals since 2019 under banking law, but the AI Act adds an extra layer of protection in cases where a fully automated algorithm made the decision.

A customer whose application was rejected by a creditworthiness assessment system can request that a human analyst review the case again, and can additionally demand an explanation of the algorithm's role in the decision. Detailed obligations for high-risk systems in certain categories, including some scoring mechanisms, will only take full effect from December 2027, meaning the rules will be phased in gradually over the coming months.

The Polish Perspective

The approaching deadline of August 2, 2026, when further AI Act provisions take effect, including the requirement to label AI-generated content and disclose conversations with a bot, is increasing pressure on companies and institutions to prepare procedures for responding to explanation requests. The head of Poland's Personal Data Protection Office (Urząd Ochrony Danych Osobowych, UODO) has already called for urgent regulation of AI use in recruitment, pointing to a gap between the pace of algorithm deployment and the readiness of national law to oversee it.

Public institutions in Poland, including the Social Insurance Institution (Zakład Ubezpieczeń Społecznych, ZUS), use analytical tools based on predictive models, for example when analyzing medical certificates, where a system assigns score values flagging irregularities that are then taken into account by medical assessors. ZUS maintains, however, that it does not consider this tool an AI system within the meaning of the AI Act, which shows how easily institutions can try to exempt their systems from the new explainability obligations.

What's Next

Lawyers specializing in the AI Act say the coming months will show how companies and public bodies respond in practice to explanation requests, especially since many Polish institutions admit they are not certain whether the systems they use even fall under the new rules. For citizens, this means it is worth asking directly about the algorithm's role in any unfavorable decision on a job, loan, or benefit; enforcing this right depends largely on the awareness of those affected themselves.

Sources: Dziennik.pl, Forsal.pl, RP.pl, Prawo.pl