What is it about?
Modern artificial intelligence (AI) systems have achieved remarkable performance, but their success increasingly depends on the quality of the data they rely on. This paper introduces and discusses the Data‑Centric AI Manifesto, offering a contemporary vision of AI in which data—rather than algorithms alone—plays a central role. The authors analyze how key data-related issues such as quality, representativeness, consistency, bias, and data governance critically affect the performance, reliability, and trustworthiness of AI systems. The paper presents a shared conceptual framework that supports both academic research and industrial practice, encouraging a shift toward a data‑centric perspective across the entire AI lifecycle.
Featured Image
Photo by Roman Budnikov on Unsplash
Why is it important?
1) Shifts the focus of AI development from models to data, addressing real-world limitations of modern AI systems 2) Contributes to more robust, transparent, and trustworthy AI, by tackling bias and errors at their source 3) Provides a common reference framework for researchers, industry, and policymakers 4) Aligns with European priorities on responsible AI, data governance, and quality‑driven innovation
Perspectives
This article is the outcome of a collective effort carried out within the Transversal Project 7 (TP7) – Data‑Centric AI and Infrastructures of the FAIR – Future Artificial Intelligence Research extended partnership. As co‑authors, we see this work as a concrete expression of a shared vision that emerged from sustained collaboration across disciplines and institutions. Our joint contribution reflects FAIR’s mission of implementing the Italian Strategic Agenda for Artificial Intelligence, adopted in November 2021, by translating high‑level principles into actionable research directions. Within TP7, we had the opportunity to jointly reflect on how data quality, governance, and infrastructures must be repositioned at the core of AI research and innovation. FAIR, as a large‑scale PNRR‑funded project (114 million euros), provided not only resources but also an institutional and scientific framework that enabled this dialogue to take place at scale. We hope this paper demonstrates how coordinated national research initiatives can foster shared understanding and long‑term impact, and that it stimulates further discussion on data‑centric AI both within and beyond the FAIR community.
Prof. Donato Malerba
Universita degli Studi di Bari Aldo Moro
Read the Original
This page is a summary of: Data-Centric AI Manifesto: How Data Quality Drives Modern AI, Electronics, May 2026, MDPI AG,
DOI: 10.3390/electronics15091913.
You can read the full text:
Contributors
The following have contributed to this page
