Author: Ms. Milica Jevtić, Scientific and Administrative Officer at the European Public Law Organization (EPLO), Brčko Office
Abstract
The EU Artificial Intelligence Act relies on human oversight as the main safeguard for high-risk AI systems used in public administration. This article examines if oversight, as regulated by Articles 13, 14 and 26, guarantees accountability. It argues that, although the Act recognises automation bias and provides for human intervention, it does not sufficiently specify how effective oversight is to be achieved or what happens when it fails. Oversight thus remains a formal requirement rather than a real guarantee. The article recommends clearer rules on overseers’ disagreement with the system and on the explanation of decisions.
Read the full article below.
This article was written as part of the EGPL Young Fellows Program 2026, an initiative of the European Public Law Organization (EPLO) linked to the Annual Reunion of the European Group of Public Law (EGPL), held from 11 to 13 September 2026 at the EPLO campus in Legrena–Cape Sounion, Greece. Selected early-career researchers in law and governance took part in the Reunion. Each then produced an original paper on this year’s theme, “New Challenges for the State”, with guidance from mentors in the ELGU academic network. The Fellows’ papers are published as a series on the ELG Blog.