Artificial Intelligence in Asylum Procedures in the European Union: International Legal Limits of Digitalisation

The article analyzes the legal limits of the use of artificial intelligence (AI) systems in the examination of asylum applications in the European Union. The main subject is the AI systems themselves. At the same time, some of the digital tools used in the procedure (data extraction programs from electronic devices, algorithms for comparing records in databases) are not formally classified as AI, but functionally affect the outcome of a case in a similar way and therefore are also included in the analysis. The study examines under what conditions such tools perform an auxiliary function and at what point they begin to actually predetermine the assessment of the circumstances of the case, creating a risk of violating the principle of non-refoulement. The article examines dialect recognition systems, biometric identification, mobile device data analysis programs, automatic classification of applications, and risk profiling of asylum seekers. Formal-legal and comparative-legal methods are applied. The analysis is based on the case law of the Court of Justice of the European Union and the European Court of Human Rights. The article substantiates that the admissibility of using AI in this area is determined by three groups of limitations: substantive (AI is not entitled to replace individual assessment with a statistical model), procedural (the role of the system must be open and contestable), and informational-legal (the processing of personal and biometric data must be lawful and proportionate). The particular strictness of these limitations is due to the principle of non-refoulement: a wrongful refusal of protection may lead to expulsion to a place where a person faces real danger, and such an error is irreversible. Therefore, the requirements for the use of AI in this area are higher than in ordinary administrative procedures. The AI Act establishes enhanced requirements for such systems, but its potential is realized only in combination with Directive 2013/32/EU, the data protection regime, and judicial practice. The conclusions can be used in assessing the legality of introducing AI into national asylum procedures, as well as in reforming EU legislation.