Answers are grounded in this paper and cite their sources.
Research on artificial intelligence (AI) and democracy and on democratic backsliding has developed largely in parallel, leaving the constitutional control of AI in politically sensitive settings under-theorised. This single-case study examines Hungary in 2025–2026, spanning late-stage backsliding and the governmental transition after the April 2026 election, through three AI-mediated pathways: synthetic political content, platform-mediated electoral communication and state facial-image identification. Combining doctrinal analysis of statutes, European Union (EU) law and Constitutional Court decisions with official and independent campaign monitoring, it proposes a framework of three control gaps—specification, review and timing—and applies it in a structured, focused comparison of constitutional control functions. Control failed at different points. Synthetic content was governed by poorly specified rules, complaint routes closed without reasoned decisions on the merits, and EU transparency duties applied only after the vote. Platform-mediated communication received rapid review; the Constitutional Court recognised algorithmic curation as relevant but developed no accountability doctrine. Facial identification was extended to all misdemeanours without prior independent authorisation, no judicial review was sought, and its operation in practice remains unverified. The facial-identification provisions remained unchanged under the new government, raising, but not resolving, the question whether some deficits are structural.
No takes yet. Share an insight, caveat, or question.
Péter Váczi (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: