FAIR principles make scientific data more reusable. But who pays for the work, maintains the infrastructure, and sustains the people who make reuse possible?
The FAIR principles (Findable, Accessible, Interoperable, and Reusable) have shaped how researchers think about data stewardship. Yet data does not become reusable simply because it is deposited in a repository. It must be documented, cleaned, standardised, and maintained. That work requires time, expertise, and resources. As artificial intelligence (AI) increases demand for reliable, well-structured datasets, a question deserves greater attention: who pays for making data reausable?
FAIR was never intended to solve every challenge in research funding, governance, or sustainability. The problem is not the framework itself, but the gap between expectations for FAIR data and the resources available to implement it.
Data reuse takes work
Preparing data for reuse can involve resolving inconsistencies, completing metadata, documenting methods, applying standards, and addressing privacy or access restrictions. Repositories also need infrastructure and personnel to preserve data and keep it accessible.
These activities carry real costs. A 2026 study, The True Costs of Data Management in the Era of Data Sharing Mandates, illustrates the financial burden associated with preparing and sharing data in complex biomedical research settings. Although costs vary across projects, the distinction is important: producing data and preparing it for reuse are separate investments.
I saw this in practice through Bioinformatics Outreach Nigeria (BON). In 2025, we organised a training on Open Data Editor (ODE) for 30 life scientists from Nigeria and other African countries. Participants worked on identifying data inconsistencies, improving metadata, preparing datasets, and assessing FAIR compliance.
The exercise made the work behind reusable data tangible. Inconsistent labels, incomplete metadata, and unclear variables can make a dataset difficult to interpret, even when the file itself is publicly available. Addressing these problems requires appropriate tools, practical skills, and an understanding of the research context. ODE provided participants with an accessible, open-source tool for working with tabular data. But tools alone are not enough. Researchers also need training and time to apply these practices, while institutions must support the infrastructure and expertise required for long-term stewardship. This was a training initiative, not a financial cost study. Its relevance lies in showing what data preparation involves, and why that work should not be treated as an incidental responsibility.
AI raises the stakes
AI increases the importance of reliable data. Poor documentation, inconsistent labels, and unclear provenance undermine model development, evaluation, and reproducibility. Preparing data for AI requires additional checks for quality, representativeness, privacy, and suitability for the intended application.
FAIR practices help, but compliance does not guarantee that data is unbiased, ethically appropriate, or fit for every purpose. Making data accessible does not mean sensitive health information should be available without guardrails. The challenge is not to produce more data, but to ensure that it can be used responsibly. This requires continued investment in curation, validation, governance, and the people who perform this work.
There is also an environmental cost to consider. The International Energy Agency estimates that data centres consumed around 415 terawatt-hours of electricity globally in 2024. While this figure is not specific to research data, it highlights the resource demands of the digital infrastructure increasingly supporting AI and data-intensive science. Responsible stewardship must balance preservation and access with proportionate use of storage and computing resources.
Who bears the cost?
The capacity to implement FAIR practices is unevenly distributed. Well-funded institutions have data managers, reliable infrastructure, and specialist support. Researchers in under-resourced settings face limited connectivity, insufficient storage, and fewer opportunities for training for this.
Through BON, I have seen the value of making open science tools and skills more accessible. But community-led training cannot substitute for institutional investment. When data stewardship is not adequately funded, the burden falls on individual researchers, support staff, and initiatives already operating with limited resources. This creates a risk that increasingly demanding data-sharing expectations will widen existing inequalities in research participation. Data standards enable collaboration, but researchers need realistic opportunities to meet them.
What needs to change?
Three actions would make FAIR implementation more sustainable:
First, fund data stewardship as part of research. Grant budgets should account for data cleaning, documentation, curation, repository deposition, and preservation. Data management plans should identify who will maintain datasets after funding ends and how that work will be financed.
Second, recognise the people behind reusable data. Data stewards, librarians, research software engineers, and other research support professionals need appropriate roles, training, and career recognition. Researchers also need protected time to prepare and document their data.
Third, invest in equitable and sustainable infrastructure. Funders and institutions should support accessible tools, shared repositories, and practical training, particularly where resources are limited. They should also consider long-term maintenance, responsible access, and the environmental implications of digital infrastructure.
These measures wouldn’t replace FAIR. They would support the conditions needed to implement it effectively.
FAIR needs sustained investment
The growth of AI has made the value of reliable data more visible, but the work required to produce it has always been part of scientific research. Our experience with BON's ODE training showed that making data reusable involves practical work that requires skills, tools, and institutional support.
FAIR provides a framework for what good data stewardship should achieve. It does not guarantee the resources needed to achieve it.
If the research community expects data to remain reusable, it must invest in the people, tools, and infrastructure that make reuse possible. Otherwise, the costs will continue to fall on those least equipped to absorb them, and the promise of open science will remain unevenly realised.
Copyright © 2026 Seun Olufemi. Distributed under the terms of the Creative Commons Attribution 4.0 License.
