Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 31 of the 31 references it cites were found in a registry or an index; 22 of the 67 sources its claims cite were checked for support.

Human-centric AI governance in the European Union. Accountability, fundamental rights, and institutional resilience in public administration

Open access
Dragomir, Andreea Nicoleta, Bulea, Iulia, Tarnu, Lucian Ioan · 2026 · Frontiers in Political Science

Abstract

Artificial intelligence is increasingly used in public administration to classify individuals, assess risks, prioritize cases, support eligibility determinations and guide the allocation of public resources. In the European Union, these uses are governed by the Artificial Intelligence Act, the GDPR and the Charter of Fundamental Rights. Formal compliance, however, does not by itself ensure lawful and accountable administration. AI relocates discretion from the visible act of decision-making to less visible choices concerning data, model design, procurement, thresholds and interface architecture. This Policy and Practice Review therefore treats human-centric AI governance not as a general ethical aspiration, but as an administrative and constitutional framework for governing public power. Drawing on EU law, public administration scholarship and a comparative institutional analysis of selected Member State practices, it develops six interdependent dimensions: legal anchoring, accountable discretion, fundamental rights by design, meaningful human oversight, contestability and justification, and institutional resilience. The analysis shows that common EU rules may produce unequal levels of protection where public authorities differ in technical expertise, audit capacity, procurement independence and access to effective remedies. It also argues that accountability must follow the chain of influence through which algorithmic systems shape administrative outcomes, rather than only the formal chain of decision-making. The article translates this framework into actor-specific recommendations concerning fundamental rights impact assessments, procurement, auditability, human oversight, transparency, contestability and post-deployment monitoring. It concludes that AI-enabled administration remains legitimate only where public authorities retain the capacity to understand, justify, correct, suspend and democratically control the systems they use.

Overview

CoChat Summary

TL;DR

This Policy and Practice Review argues that human-centric AI governance in EU public administration must be treated as an administrative and constitutional framework for governing public power, not a general ethical aspiration. It develops a six-dimensional framework and shows, through a comparative reading of the Netherlands, France and Estonia, that harmonized EU rules can produce unequal protection where authorities differ in expertise, audit capacity, procurement independence and access to remedies.

Key findings

  • AI does not eliminate administrative discretion but relocates it upstream into data selection, model design, threshold setting, procurement and interface architecture; accountability must therefore follow the "chain of influence" rather than only the formal chain of decision-making.
  • The article proposes six interdependent dimensions: legal anchoring, accountable discretion, fundamental rights by design, meaningful human oversight, contestability and justification, and institutional resilience (Table 2), each tied to distinct legal anchors (AI Act, GDPR, Charter Articles 41 and 47, NIS2 where applicable) and institutional safeguards.
  • The comparative analysis locates safeguards at different stages: the Netherlands relies on ex post judicial correction (SyRI, Rechtbank Den Haag 2020, finding the legislation failed the fair balance under Article 8(2) ECHR); France embeds statutory procedural transparency in administrative law (Articles L311-3-1 and R311-3-1-1 of the Code des relations entre le public et l'administration); Estonia illustrates coordinated digital-government capacity and shared infrastructure (OECD 2026a), which the authors stress is an enabling condition rather than a substitute for rights safeguards.
  • The paper argues that formal compliance with the AI Act's risk-based architecture (Articles 9–15, 14, 27, 71, 86) and Article 22 GDPR cannot by itself secure legitimate administrative use, because risk is both technological and institutional; nominal human involvement may reduce the apparent relevance of Article 22 GDPR even where algorithmic output retains decisive influence (citing Sarra, 2025).

Methods

  • Design: structured legal-doctrinal and comparative governance approach; the article explicitly does not measure AI adoption levels or assess technical performance of individual systems.
  • Corpus: primary legal texts (AI Act, GDPR, Charter of Fundamental Rights, relevant national administrative law and judicial decisions, with NIS2 considered only for the institutional-resilience dimension) plus policy documents (2019 Ethics Guidelines for Trustworthy AI, EU implementation materials) and secondary literature identified via targeted searches in Scopus, Web of Science, HeinOnline and Google Scholar, with citation tracing, covering materials up to 20 July 2026.
  • Comparative component: purposive, maximum-variation selection of the Netherlands, France and Estonia, examined against a common grid (legal and procedural safeguards, location of oversight, administrative and technical capacity, provider dependence, transparency and contestability, capacity for correction and learning); analysis proceeds through four operations — identifying legal duties, tracing discretion across the process, examining the three settings against the grid, and cross-case comparison feeding the six-dimensional framework.

Limitations

  • Stated in the paper (Section 12): the study relies on legal, policy and institutional documents and does not assess technical performance or measurable effects of individual AI systems; the three national settings were selected for contrasting features and do not support a general ranking of Member States; documentary evidence varies across jurisdictions and several AI Act mechanisms remain in development; the framework supports qualitative institutional inference rather than causal generalization, and its practical value should be tested through further empirical research on procurement, FRIAs, human oversight, administrative remedies and post-deployment monitoring.
  • Evident from the text: the comparative cases are not treated as representative of all algorithmic use within their administrations, and the authors note that SyRI's precise technical characteristics were not fully public and the judgment did not establish a general prohibition on automated fraud detection (citing Van Bekkum and Zuiderveen Borgesius, 2021).
  • Evident from the text: the article is a Policy and Practice Review whose recommendations are derived from documentary and comparative inference; no primary empirical data (e.g., interviews, procurement records, audit results) are collected or analyzed.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3389/fpos.2026.1928481

Cites

31 of 31 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

22 Valid · 3 Not valid · 43 Not checked · 176 Own results and interpretations

119 more claims · 6 with their cited sources checked

Create a free account to see all 244 claims and to verify the 2 references still open.

See all claims — free account
  • It may state two separate claims as one. 83
  • Probably not a substantial claim. 68
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Citations

Cites

Show all 30 references

Research works printed but not linked: 5

  1. 2012, Charter of fundamental rights of the European Union, Off. J. Eur. Union C, 326, 391
  2. 2016, Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation), Off. J. Eur. Union L, 119, 1
  3. 2022, Directive (EU) 2022/2555 of the European parliament and of the council of 14 December 2022 on measures for a high common level of cybersecurity across the Union (NIS 2 Directive), Off. J. Eur. Union L, 333, 80
  4. 2024, Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), Off. J. Eur. Union L
  5. Digital Government Outlook 2026: From Foundations to Transformational Impact

Metadata

Authors
Dragomir, A. N., Bulea, I., & Tarnu, L. I.
Affiliations
Lucian Blaga University of Sibiu (RO)
Corresponding author
Dragomir, Andreea Nicoleta
Venue
Frontiers in Political Science
Published
2026-08-14
Volume
8
ISSN
2673-3145 (ISSN-L)
OpenAlex
W7203459260
Cited by
1
References
30
Citations by year
2026: 1
FWCI
34.24
Type
journal article
Language
en
Open access
Locations
  • Frontiers in Political Science (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Topics
  • Social Sciences › Law › Digitalization, Law, and Regulation
  • Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
  • Social Sciences › Law › Administrative Law and Governance
Keywords
AI governance, public administration, European Union, EU AI Act, GDPR, Charter of Fundamental Rights, accountability, meaningful human control, algorithmic decision-making, administrative discretion, contestability, institutional resilience, fundamental rights impact assessment, auditability, transparency, post-deployment monitoring, effective remedy, EU member states

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Human-centric AI governance in the European Union. Accountability, fundamental rights, and institutional resilience in public administration. https://app.cochat.ai/papers/10.3389/fpos.2026.1928481
The paper
Dragomir, A. N., Bulea, I., & Tarnu, L. I. (2026). Human-centric AI governance in the European Union. Accountability, fundamental rights, and institutional resilience in public administration. Frontiers in Political Science. https://doi.org/10.3389/fpos.2026.1928481

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.