Authors: Alessia Borromeo (Senior Researcher GEO Politica) and Tommaso Ravizza (Autore - area tematica Organizzazioni Internazionali MI Post)
Abstract: Artificial intelligence is gradually becoming part of European asylum procedures, supporting activities such as translation, transcription, information retrieval and case management. Yet these technologies operate in a context where relevant information is not always available from the outset: vulnerabilities linked to trafficking, violence or psychological distress may emerge over time, while digital systems rely on information that can be recorded, organised and made available for administrative use. Situating AI within the broader digitalisation of European asylum governance, this analysis examines how technological applications, Eurodac and database interoperability contribute to the administrative representation of individual cases. It argues that greater administrative visibility does not necessarily amount to a more complete understanding of applicants’ circumstances, and asks how digital systems can accommodate evolving information and support its interpretation throughout asylum procedures.
Introduction: AI and the Digitalisation of European Asylum Procedures
Discussions of artificial intelligence in asylum procedures often concentrate on the final stage of decision-making, particularly on the possibility that algorithms may help determine eligibility for entry, residence or international protection. This focus can obscure developments already under way at earlier stages, where AI is increasingly used for translation, transcription, information retrieval and case management. Some applications also contribute to producing or verifying information that may later inform the assessment of an individual claim (Dumbrava, 2025; Ozkul, 2023; EUAA, 2026).
The growing use of AI forms part of a wider process of digitalisation. In 2023, the European Union Agency for Asylum (EUAA) introduced a strategy promoting digital innovation in asylum procedures and reception systems across EU+ countries (EUAA, 2023). In January 2026, the European Commission gave technological development a more explicit role in its long-term approach to migration governance: its European Asylum and Migration Management Strategy presents digitalisation and artificial intelligence as instruments for improving the quality, consistency and timeliness of administrative procedures. It envisages integrated case-management systems connecting screening, asylum, reception and transfer procedures, alongside possible uses of AI in interpretation, translation, transcription, Country of Origin Information, workflow management and preliminary checks of application completeness (European Commission, 2026a).
In this context, digital tools may reduce administrative burdens, facilitate access to relevant material and improve coordination across procedural stages. Their significance, however, extends beyond operational efficiency: the systems used to process and retrieve information also shape the material available to authorities when they assess individual applications. Asylum procedures are particularly sensitive to this dynamic because the circumstances underlying an application may not be fully known at the outset. Personal experiences can be difficult to communicate, their significance may become apparent only later, and relevant information can emerge progressively through interviews and subsequent interactions (Skrifvars et al., 2025; Nicodemi and Cirillo, 2024). The question is therefore not simply who ultimately decides, but what the administration is able to recognise when assessing a case, and how that understanding develops throughout the procedure.
This article first examines the kinds of AI applications entering asylum procedures and the difficulty of representing vulnerability administratively, also turning to data infrastructures such as Eurodac and interoperability, and finally considering system design and human oversight as conditions for revising earlier assessments.
AI, Vulnerability and Administrative Representation
Artificial intelligence enters asylum procedures through technologies performing different functions. Some primarily support administrative activities, while others interact more directly with information relevant to assessing individual applications (Dumbrava, 2025; Ozkul, 2023). The Asylum Report 2026 illustrates this diversity: Austria has been developing speech-to-text and translation tools for asylum interviews; Belgium has tested an AI-powered search system to retrieve information relevant to appeal decisions and has introduced AI into Country of Origin Information activities; and France has piloted AI-assisted document searches, while excluding their use for drafting decisions (EUAA, 2026). Dumbrava (2025) also identifies applications involving dialect recognition, name transliteration, automatic transcription and case matching, complementing Ozkul’s (2023) earlier mapping of digital technologies in European migration and asylum governance.
The implications of these applications depend on their functions and on the stages at which they interact with information. Transcription tools influence how interviews are recorded, search systems facilitate access to existing material, and technologies used to establish or verify geographical origin may contribute more directly to the substantive assessment of a claim. These differences matter most when the information concerns applicants’ vulnerabilities, whose identification may depend on circumstances that are neither immediately apparent nor easily captured through standardised administrative processes. The IOM Determinants of Migrant Vulnerability framework understands vulnerability as resulting from the interaction of risk and protective factors at the individual, household or family, community and structural levels (IOM, 2019). Individual circumstances are thus shaped by wider social and institutional environments, and needs and exposure to harm may change over time or become visible only through continued interaction.
While the IOM framework offers a broader understanding of vulnerability, EU asylum legislation translates certain individual circumstances into specific procedural and reception-related obligations. Regulation (EU) 2024/1348, which applies from June 2026 as part of the Pact on Migration and Asylum, provides for identifying applicants who require special procedural guarantees, while Directive (EU) 2024/1346 requires special reception needs to be assessed as early as possible, recognising that additional needs may emerge later and providing for their continued monitoring throughout the procedure (European Parliament and Council of the European Union, 2024a; 2024b). The EUAA’s operational standards on vulnerability similarly treat assessment as an ongoing responsibility rather than an isolated administrative event (EUAA, 2025). Empirical research in Finland also illustrates the difficulties of communicating sensitive experiences during asylum interviews (Skrifvars et al., 2025). Nicodemi and Cirillo (2024) further show that the categories of asylum seeker, refugee and victim of trafficking may overlap without adequately reflecting individual trajectories, particularly when experiences of coercion, violence or exploitation become visible only progressively.
The difficulties involved in representing complex personal circumstances predate the introduction of artificial intelligence: asylum administrations have long relied on interviews, documents and administrative categories to assess individual experiences. Digitalisation nevertheless changes the conditions under which information is recorded and made available, since technological systems increasingly take part in processes that influence how applicants are represented throughout the procedure.
From Case Files to Data Infrastructures
The transformation of Eurodac illustrates the growing role of information infrastructures in European migration governance. Originally established as a fingerprint database supporting the Dublin system, Eurodac now serves a broader range of administrative purposes under its recast Regulation (EU) 2024/1358 (European Parliament and Council of the European Union, 2024d). According to the Commission, the new Eurodac became operational in June 2026, and the European Search Portal and the Common Identity Repository are now operational as well. The latter allows identity data held in different systems to be cross-checked, while the former gives authorised users a common interface for searching several EU information systems (European Commission, 2026b).
Eurodac and interoperability perform different functions from artificial intelligence, but they belong to the same environment, in which migration is increasingly administered through structured and interconnected information. Bellanova and Glouftsios (2022) argue that EU interoperability does not create a single unified database but reorganises and repurposes existing datasets through shared mechanisms and standards. Their concept of “formatting” draws attention to how information infrastructures shape mobility governance by determining the conditions under which data are structured, connected and made available for institutional use. Trauttmansdorff and Loschi (2025) similarly emphasise the administrative, legal and political dimensions of interoperability, showing how technological arrangements reflect different approaches to managing mobility.
Datafication is thus reshaping how individual asylum cases become administratively visible, by translating identities, movements and procedural histories into structured data that institutions can store, search and connect. Greater accessibility can facilitate coordination and retrieval, but the value of the resulting records depends on their quality, representativeness and interpretative context; limitations in these dimensions can distort analyses and produce incomplete representations of complex individual circumstances (Borromeo, 2026). The point is especially relevant in asylum governance, where standardised administrative information coexists with personal experiences that require disclosure or contextual assessment. Biometric identifiers and previous registrations, for example, may be readily accessible, while experiences of trafficking, psychological distress or other vulnerabilities remain undocumented or acquire significance only later. Different information systems serve different institutional purposes, and their capacity to identify individuals does not necessarily extend to understanding the circumstances relevant to their protection needs.
How administrative information is organised may also affect which aspects of an applicant’s situation enter the assessment. Standardised records are generally easier to retrieve and compare, whereas information about personal experiences may require further interaction and professional interpretation. Digital infrastructures supporting asylum procedures should therefore be judged not only by what they can retrieve, but by how they integrate standardised records with information that emerges through continued interaction.
From Information Design to Human Oversight
If administrative visibility depends partly on how information is structured, system design becomes an important dimension of asylum governance. Standardisation is necessary for authorities managing complex procedures, providing shared categories, retrievable records and continuity across stages. At the same time, administrative representations need to remain open to revision when new circumstances emerge or previously recorded information acquires a different meaning. Information collected at the beginning of a case may carry particular weight once it has been classified and incorporated into administrative records, while later disclosures can remain dispersed across interviews, documents or communications: integrating them may require reconsidering earlier classifications. The Slovak MIGRA system offers a useful example: in 2025, two specialised sub-files were introduced into applicant files, one dedicated to vulnerabilities and another to social documentation (EUAA, 2026). The available evidence does not establish whether this change has improved vulnerability identification in practice, but it illustrates how information architecture can create dedicated spaces for material that might otherwise remain fragmented.
Two conditions therefore shape the visibility of vulnerability. The first is system design: whether administrative records can be updated, previous entries contextualised and subsequent disclosures made accessible throughout the procedure. The second is effective human oversight, which allows available information to be critically interpreted and earlier assessments to be reconsidered as new circumstances emerge. The EU AI Act provides a regulatory framework for the latter. Annex III classifies several applications in migration, asylum and border control as high-risk, including systems intended to assist competent authorities in examining asylum, visa and residence-permit applications, including assessing the reliability of evidence. Article 14 requires high-risk AI systems to be designed to enable effective human oversight, allowing those responsible to understand their capabilities and limitations, interpret outputs appropriately and remain aware of automation bias (European Parliament and Council of the European Union, 2024c).
In asylum procedures, however, meaningful oversight also concerns earlier stages of information processing. By the time an official receives an automated output, relevant material may already have been translated, transcribed, classified or retrieved through technological systems. The assessment then relies on a representation shaped by preceding administrative and technical processes, whose limitations may not be apparent at the point of decision. Moreover, support functions such as transcription, translation or document search may not qualify as high-risk under Annex III, and may therefore fall outside Article 14. Oversight should accordingly extend to the reliability of the information on which assessments depend, including the distinction between recorded information and technological inferences, the possibility of correcting administrative records and officials’ capacity to recognise the limitations of the systems involved. This goes beyond the specific requirements of Article 14 to the broader governance arrangements through which information is managed in asylum procedures. Transparency, accountability, shared standards and adequate professional competencies support administrations’ capacity to interpret available information and to revise earlier assessments when new circumstances require it.
The practical implementation of these arrangements remains a matter for empirical investigation. In 2026, the EU Agency for Fundamental Rights launched a research project on AI-powered technologies used in asylum, visa, residence-permit and return procedures, with particular attention to their fundamental-rights implications and the safeguards accompanying their deployment (FRA, 2026). Further evidence on how these applications operate in different administrative contexts will be important for understanding the relationship between technological development and effective protection safeguards.
Conclusions: From Administrative Visibility to Meaningful Knowledge
Artificial intelligence and interconnected data infrastructures are changing how information is managed within European asylum procedures. Digitalisation can improve administrative coordination and facilitate access to relevant records, but greater visibility does not necessarily bring a more complete understanding of applicants’ circumstances. Experiences of trafficking, violence or psychological distress may remain undisclosed during initial assessments, while previously recorded information can acquire new significance as individual situations become clearer. The effectiveness of digital systems therefore depends also on their capacity to preserve context, incorporate subsequent disclosures and support the revision of earlier assessments.
Assessing digital systems in asylum procedures thus requires attention to this capacity, alongside technical performance and administrative efficiency. Human oversight plays an important role here, but retaining authority over final decisions, though necessary, cannot fully address limitations that arise as information is produced, organised and retrieved. The governance of AI in asylum procedures is consequently closely connected to the broader governance of information, including the institutional arrangements through which administrative representations are constructed and reconsidered.
The central political challenge is to ensure that growing administrative capacity does not lead institutions to mistake the visibility produced by digital infrastructures for comprehensive knowledge of the individuals concerned. Ultimately, the question for European asylum governance is not simply who becomes visible through digital systems, but what becomes visible, how that representation is constructed, and whether institutions remain capable of recognising and responding to what their existing categories do not adequately capture.
The analysis builds on questions first explored in Episode 4 of Season 2 of the GeopoliticaMI podcast, “Migrazioni nell’era dell’IA: dati, vulnerabilità e governance europea”.
References
Bellanova R., Glouftsios G. (2022), “Formatting European security integration through database interoperability”, European Security, vol. 31, n. 3, pp. 454-474, DOI: 10.1080/09662839.2022.2101886. (A-2). Available at https://www.tandfonline.com/doi/full/10.1080/09662839.2022.2101886
Borromeo A. (2026), “Dati, intelligenza artificiale e welfare: il ruolo della data philanthropy”, Quaderni di Economia Sociale, n. 1/2026, SRM, ISSN 2421-0315. (A-2). Available at https://www.sr-m.it/it/cat/prod/322891/quaderni-di-economia-sociale-1-2026.htm
Dumbrava C. (2025), Artificial intelligence in asylum procedures in the EU, European Parliamentary Research Service, European Parliament, Briefing, 7 July 2025. (A-2). Available at https://www.europarl.europa.eu/thinktank/en/document/EPRS_BRI(2025)775861
European Commission (2026a), European Asylum and Migration Management Strategy, COM(2026) 45 final, Brussels, 29 January 2026. (A-1). Available at https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex:52026DC0045
European Commission (2026b), Digital border and migration management in the EU, Directorate-General for Migration and Home Affairs, 26 June 2026. (A-1). Available at https://home-affairs.ec.europa.eu/policies/migration-and-asylum/migration-management-eu-member-states/digital-border-and-migration-management-eu_en
European Parliament, Council of the European Union (2024a), Regulation (EU) 2024/1348 of 14 May 2024 establishing a common procedure for international protection in the Union and repealing Directive 2013/32/EU, Official Journal of the European Union. (A-1). Available at https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A02024R1348-20260227
European Parliament, Council of the European Union (2024b), Directive (EU) 2024/1346 of 14 May 2024 laying down standards for the reception of applicants for international protection, Official Journal of the European Union. (A-1). Available at https://eur-lex.europa.eu/eli/dir/2024/1346/oj/eng
European Parliament, Council of the European Union (2024c), Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), Official Journal of the European Union. (A-1). Available at https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
European Parliament, Council of the European Union (2024d), Regulation (EU) 2024/1358 of 14 May 2024 on the establishment of ‘Eurodac’ for the comparison of biometric data (recast), Official Journal of the European Union. (A-1). Available at https://eur-lex.europa.eu/eli/reg/2024/1358/oj/eng
European Union Agency for Asylum – EUAA (2023), EUAA Strategy on Digital Innovation in Asylum Procedures and Reception Systems, September 2023. (A-1). Available at https://www.euaa.europa.eu/publications/euaa-strategy-digital-innovation-asylum-procedures-and-reception-systems
European Union Agency for Asylum – EUAA (2025), Operational Standards and Indicators on Vulnerability-related Aspects in the Asylum Procedure, November 2025. (A-1). Available at https://www.euaa.europa.eu/publications/operational-standards-indicators-vulnerability-aspects-asylum-procedure
European Union Agency for Asylum – EUAA (2026), Asylum Report 2026, Publications Office of the European Union, Luxembourg, DOI: 10.2847/0813257. (A-2). Available at https://www.euaa.europa.eu/asylum-report-2026
European Union Agency for Fundamental Rights – FRA (2026), Use of artificial intelligence in asylum and immigration procedures – fundamental rights implications, ongoing research project, 20 February 2026. (A-1). Available at https://fra.europa.eu/it/project/2026/use-artificial-intelligence-asylum-and-immigration-procedures-fundamental-rights
International Organization for Migration – IOM (2019), IOM Handbook on Protection and Assistance for Migrants Vulnerable to Violence, Exploitation and Abuse, IOM, Geneva. (A-1). Available at https://publications.iom.int/books/iom-handbook-migrants-vulnerable-violence-exploitation-and-abuse
Nicodemi F., Cirillo C. (2024), “Toward effective protection of victims of human trafficking in mixed migration flows: referral mechanisms shaped on individual need. The Italian experience and the European perspective”, Frontiers in Human Dynamics, vol. 6, 1436612, DOI: 10.3389/fhumd.2024.1436612. (A-2). Available at https://www.frontiersin.org/journals/human-dynamics/articles/10.3389/fhumd.2024.1436612/full
Ozkul D. (2023), Automating Immigration and Asylum: The Uses of New Technologies in Migration and Asylum Governance in Europe, Refugee Studies Centre, University of Oxford. (A-2). Available at https://www.researchgate.net/publication/367350622_Automating_Immigration_and_Asylum_The_Uses_of_New_Technologies_in_Migration_and_Asylum_Governance_in_Europe
Skrifvars J., Ilmoni A., Siegfrids L., Galán M., Selim H., Stevens L., Korkman J., Antfolk J. (2025), “Experiences of asylum interviews by asylum officials, interpreters and asylum seekers in Finland”, Journal of Criminal Psychology, vol. 15, n. 2, pp. 260-280, DOI: 10.1108/JCP-09-2024-0085. (A-2). Available at https://research.abo.fi/en/publications/experiences-of-asylum-interviews-by-asylum-officials-interpreters/
Trauttmansdorff P., Loschi C. (2025), “Interoperability and the multiple modes of ordering in Europe’s digital border regime”, International Migration, vol. 63, n. 3, e70022, DOI: 10.1111/imig.70022. (A-2). Available at https://onlinelibrary.wiley.com/doi/full/10.1111/imig.70022
Cover photo by Matteo Vontz on Unsplash
Classification of sources and information:
1 | Confirmed | Confirmed by other independent sources; logical in itself; coherent with other information on the topic |
|---|---|---|
2 | Presumably true | Not confirmed; logical in itself; coherent with other information on the topic |
3 | Maybe true | Not confirmed; reasonably logical in itself; coherent with some other information on the topic |
4 | Uncertain | Not confirmed; possible but not logical in itself; no other information on the topic |
5 | Improbable | Not confirmed; not logical in itself; contradicts with other information on the topic |
6 | Not able to be evaluated | No basis to evaluate the validity of the information |
Trustworthiness of the source | ||
A | Trustworthy | No doubt about authenticity, reliability or competence; has a history of total trustworthiness |
B | Normally trustworthy | Small doubts about authenticity, reliability or competence, nevertheless has a history of valid information in a majority of cases |
C | Sufficiently trustworthy | Doubts about authenticity, reliability or competence; however, has supplied valid information in the past |
D | Normally not trustworthy | Significant doubt about authenticity, reliability or competence, however has supplied valid information in the past |
E | Not trustworthy | Lack of authenticity, reliability or competence; history of invalid information |
F | Not able to be evaluated | No basis to evaluate the validity of the information |
Riproduzione Riservata Ⓡ