Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
Responsible AI in Polish Higher-Education Administration under the EU AI Act: High-Risk Use Cases, Fundamental-Rights Impact Assessment, and Institutional Accountability
Citation. Stadniczeńko, D. (2026). Responsible AI in Polish higher-education administration under the EU AI Act: High-risk use cases, fundamental-rights impact assessment, and institutional accountability. AI, Law, Politics, 2(S1), 82–112. https://doi.org/10.5709/alp-02.s1.2026-04
Summary. This doctrinal and legal-design article maps how Polish higher-education institutions should classify and govern administrative AI under the EU AI Act, GDPR, and relevant Polish law. It proposes a use-case classification matrix, a FRIA–DPIA boundary map, a four-gate institutional model, and an accountability file spanning procurement through retirement. The framework treats meaningful human authority, intelligible reasons, contestability, accessibility, and auditable institutional ownership as conditions of responsible deployment.
Key claims
- Admissions, learning-outcome evaluation, level determination, test monitoring, and employment uses may be Annex III high-risk uses, while narrow procedural tools require fact-specific workflow assessment.
- A formal human signature does not by itself prevent material AI influence; the relevant unit of analysis is the deployed workflow, including outputs, defaults, reviewer practice, legal consequence, and available remedy.
- AI Act fundamental-rights impact assessment and GDPR data-protection impact assessment have distinct triggers and functions, so a FRIA neither applies to every private university nor replaces a DPIA.
- AI emotion inference in educational institutions is generally prohibited, apart from medical or safety purposes, and relabeling it as engagement, attention, or stress detection does not change the functional analysis.
Evidence & limitations
- The article uses doctrinal and legal-design analysis of legislation, case law, official guidance, and selected interdisciplinary scholarship available through 2026-08-11; its practical tools require future empirical validation.
- The analysis is specific to Polish higher education and an evolving EU AI Act implementation timeline, so its classifications and institutional recommendations require jurisdiction- and workflow-specific reassessment.