Your team finishes a dataset that answers a question a partner lab abroad is also studying. You share it with them. They still cannot use the file.
The reason is almost never the file format. The access terms may be unclear, or your metadata may assume background only your own lab has. The repository holding the file may lose its funding when the grant ends, or the law in your partner’s country may limit what they are allowed to store.
Each of those is a problem you can fix. None of these issues appears in a model evaluation report.
This article explains what makes research data usable across a border. You will see what the FAIR, CARE and TRUST principles each govern and what your own standards work adds. A climate assessment body supplies the worked example that follows.
What Science Diplomacy Means for Data Work
Science diplomacy is the use of scientific cooperation to support relations between countries, along with the use of diplomatic channels to keep that cooperation going.
UNESCO marks World Science Day for Peace and Development on November 10 and gives the 2026 theme as "Common ground through science diplomacy."
This cooperation is most evident in the data. Two governments can sign a climate research agreement, and nothing comes of it until the observations sit in a form both sides can find and check. Open data science therefore rests on agreements that outlast any single institution.
What Makes Data Usable Across a Border
An agreement to share data is only the starting point. The practical problems appear when someone tries to act on it. International research data belongs to institutions that answer to different funders and different laws, each with its own retention rules. Any one of those differences can stop a transfer before it starts.
Angela Liberatore of the European University Institute argues in Royal Society Open Science that "Data and data infrastructures are key for scientific work". They also form "a key part of the evidence needed for sound international negotiations on global issues."
Her paper identifies three pressures on open access to that data. States sometimes act to "forbid access, hide or destroy scientific data." Large platforms hold vast collections and are "mostly privatizing them for market purposes." Language models make data mining faster while "reproducing and possibly amplifying biases."
Access on its own therefore solves only part of the problem. A dataset that is open this year and unfunded next year was never really shared. A file nobody documented cannot be checked by anyone who was absent when the numbers were collected.
What FAIR, CARE and TRUST each govern
Three sets of principles answer those gaps, and people tend to name them together without saying how they differ. You can apply them far more easily once you see that each one covers a different object.
| Framework | What it covers | The principles | What it asks of you |
|---|---|---|---|
| FAIR | The dataset | Findability, Accessibility, Interoperability, Reusability | Publish rich metadata a machine can read and state how people get access. Use formats that follow domain standards. |
| CARE | The people behind the data | Collective benefit, Authority to control, Responsibility, Ethics | Establish who holds rights over the data and record how you will use it. Show what the community receives in return. |
| TRUST | The repository | Transparency, Responsibility, User focus, Sustainability, Technology | Write down how you preserve files and fund the repository past the project. Keep the technology fit for purpose. |
FAIR arrived in 2016 and set the working baseline for open data. Liberatore notes that open access has since become a requirement for public funding recipients in several national and European programs. Under Horizon Europe, open science is a legal obligation. Open access to publications and to research data are both mandatory practices.
CARE fills a gap in FAIR, which says nothing about consent or ownership. The Global Indigenous Data Alliance developed it so that communities described by a dataset benefit from it and keep real authority over its use. The two frameworks work together, and in practice that means asking who the data is about before deciding how to publish it.
TRUST sets conditions for the repository itself. Even a dataset anyone can find needs one that will still exist in ten years, with a written preservation practice and a funding model behind it.
How the IPCC Shares Climate Data Across Borders
Those three ideas are easier to follow in a organization that already does this work. The Intergovernmental Panel on Climate Change runs a Data Distribution Centre (DDC), which supplies the data behind its assessment reports to researchers worldwide.
A standing group oversees it, the Task Group on Data Support for Climate Change Assessments. Its terms of reference tell it to guide the DDC in support of "curation, transparency, traceability and stability of data and scenarios," and to help with "the availability and consistent use" of them.
Read that mandate as a job description and it starts to look familiar. Curation is documentation and metadata work, and traceability is lineage. Stability means deciding how long you keep a file and in what format. Consistent use means interoperability, or an agreement that a variable means the same thing whichever modeling center produced it.
That document also tells the group to look at "options for a sustainable structure, functioning, and resourcing" for the DDC. Those are TRUST concerns, written into the mandate of an intergovernmental body.
What Data Professionals Contribute
Governments agree on the intent. You build the conditions that let data move, and four areas carry that work.
- Metadata and documentation. A reader outside your institution has to act on your description alone, so write it for machines as well as for people.
- Interoperability. Shared schemas and stable identifiers enable datasets to integrate without manual record matching. This is the discipline that holds a modern data engineering stack together, applied past the edge of one organization.
- Provenance. Record how you collected the data and how you later changed it, so a partner can judge whether it fits their question.
- Repository practice. Preservation formats and succession plans decide whether a dataset survives a funding cycle.
This work is integral to ordinary data engineering. The same habits that keep a pipeline maintainable inside one organization make a dataset usable beyond it, which is why the evolving role of the data engineer now reaches into stewardship and governance.
It is important to clearly acknowledge one limitation. People often summarize the rule as research that is "as open as possible, as closed as necessary." Liberatore points out that this still leaves open "how a balance can be reached, based on what criteria, and who is involved in the decision-making."
Personal data and security rules are real constraints, as are questions of data sovereignty. Draw that line on purpose and write down where you drew it.
Conclusion
Effective science diplomacy depends on accessible and trustworthy evidence shared across borders. Good metadata and funded repositories meet that test, which places a real share of the work with data professionals.
The three principle sets split the job cleanly. FAIR settles whether anyone can use a dataset. CARE decides who may authorize that use, and TRUST determines whether the dataset will still exist when someone looks for it a decade from now.
Take one dataset your organization shares with outside partners and put those three questions to it. The answers will tell you whether you published data or simply released a file.
