Crafting data sharing agreements for international food security

Food insecurity remains one of the most stubborn development challenges of the twenty-first century, and the datasets that researchers generate are often too valuable to remain locked behind institutional firewalls. Wheat breeders in Narrabri, agronomists studying soil salinity in the Murray-Darling Basin, and remote sensing specialists analysing cropping patterns across the Mekong Delta all need access to each other's observations. Without structured arrangements, that flow of evidence slows, duplicates, or simply never arrives.

The Australian research community has long understood this. CSIRO, the Australian Centre for International Agricultural Research, and universities from Perth to Brisbane collaborate routinely with partners in Sub-Saharan Africa, South-East Asia, and the Pacific. Each collaboration carries its own legal culture, funding rules, and assumptions about who owns a soil sample, a satellite tile, or a household survey. Drafting a data sharing agreement that respects those differences while still delivering interoperable outputs is the practical heart of international food security work.

The pages that follow walk through the legal, technical, and cultural choices that shape these agreements, drawing on discussions held at events such as the IASSIST 2017 conference in Lawrence. For teams preparing a new project, the published data management plan template for digital agriculture complements the formal sharing instruments described here.

The shared data imperative for global food security

Food security is, by definition, a global question. A failed monsoon in India reshapes wheat prices in Adelaide; a new strain of rust in East African barley fields changes the disease screening protocols used in Horsham. Researchers can only respond to these cascades when their evidence base is comparable across countries, seasons, and farming systems. Data sharing agreements are the mechanism that makes comparability possible without forcing every contributor to surrender control of their information.

The imperative has sharpened in recent years. Open data mandates from major funders, including the Bill & Melinda Gates Foundation, the European Commission, and Australia's own Department of Foreign Affairs and Trade, have moved data sharing from an optional courtesy to a contractual expectation. Reviewers evaluating a grant now routinely ask whether the proposed outputs will be discoverable, reusable, and ethically stewarded long after the final report.

There is also a planetary argument. Crop models, climate projections, and yield forecasts depend on training datasets that no single country can assemble alone. Phenotyping networks that link glasshouses in Sydney with field trials in Tamale or Hanoi generate the volumes needed to train machine learning systems on real agricultural variability. When those flows are governed well, the resulting science is sharper; when they are governed poorly, trust between partners erodes.

Legal foundations and ethical commitments

A workable data sharing agreement rests on four pillars: clear ownership, permitted uses, security obligations, and a pathway for resolving disputes. Ownership clauses usually distinguish between raw data, derived data, and metadata, often assigning stewardship rather than absolute title to a designated custodian. Permitted use clauses spell out which research questions can be pursued, whether commercial spin-offs are allowed, and how attribution will appear in publications.

Ethical commitments have grown more demanding. Many agreements now embed the FAIR principles (findability, accessibility, interoperability, and reusability) alongside CARE principles for Indigenous data governance. CARE, developed through the Global Indigenous Data Alliance, foregrounds collective benefit, authority to control, responsibility, and ethics. Australian institutions increasingly reference CARE when working with Aboriginal and Torres Strait Islander communities whose traditional ecological knowledge informs bushfood research, fire management studies, and rangelands monitoring.

Security and privacy clauses deserve particular attention when personal or location-sensitive data is involved. Australian partners typically align with the Privacy Act and the Australian Government Information Security Manual, while collaborators in the European Union operate under the General Data Protection Regulation. Aligning the two regimes within a single agreement often requires a short annex that maps obligations across jurisdictions.

Governance models that travel across borders

There is no single template for governing shared data, and the best model depends on the partnership's lifespan, funding base, and trust history. Three approaches dominate the international food security landscape.

Bilateral agreements work well when two institutions run a tightly defined experiment, such as a joint wheat pre-breeding programme between the University of Adelaide and a counterpart in Morocco. They are fast to draft and easy to amend, but can leave other consortium partners exposed if their access is negotiated separately.

Multilateral consortium agreements suit larger ventures like the CGIAR system or the ARDC-funded Australian research data commons. A single master agreement covers all participants, with schedules describing each member's contributions. The trade-off is drafting time: legal teams from Canberra, Nairobi, and Jakarta must converge on shared language, often across three or more working languages.

Federated or hub-and-spoke models let each participating country retain its own data repository while agreeing on common access protocols. A central metadata catalogue points users to datasets held in Brisbane, Nairobi, or Hanoi, but the underlying files never leave their home jurisdiction. This model has gained traction among Pacific Island partners wary of uploading sensitive coastal fisheries data to overseas servers.

A brief comparison helps clarify which governance approach fits which circumstance:

Governance model	Strengths	Weaknesses	Best suited to	Typical drafting time
Bilateral agreement	Fast to negotiate, simple amendment process	Inconsistent treatment of consortium partners, duplicated effort	Two-institution joint projects	1–3 months
Multilateral consortium agreement	Consistent terms across all members, single point of reference	Long drafting phase, need for legal counsel in multiple jurisdictions	Large multi-country research programmes	6–12 months
Federated hub-and-spoke	Data sovereignty preserved, lower bandwidth requirements	Complex metadata harmonisation, governance overhead	Networks with sensitive or large datasets	3–6 months
Trust-based MoU with later formal agreement	Quick start, builds working relationships	Limited legal weight if disputes arise	Early-stage pilots and exploratory work	2–4 weeks

Technical plumbing: standards, metadata, and interoperability

Legal clauses mean little if the underlying systems cannot speak to one another. Interoperability rests on shared metadata schemas, common vocabularies, and agreed file formats. The agINFRA project, the Research Data Alliance's Agricultural Data Interest Group, and the GODAN network have produced vocabularies that translate local terms into something machines can parse. In Australia, the Australian Bureau of Statistics' data standards and the Terrestrial Ecosystem Research Network's protocols provide familiar reference points that international partners often accept readily.

Metadata should answer at least five questions for every dataset: who collected it, when, where, under what conditions, and how it can be reused. Persistent identifiers such as DOIs and ORCIDs attach that metadata to people and outputs, making citations straightforward and credit unambiguous. Many Australian institutions now use repositories supported by the Australian Research Data Commons, which provide durable identifiers and clear licensing out of the box.

Security and access infrastructure typically layers authentication, authorisation, and audit. Federated login systems like AAF's Rapid IdP and the international eduGAIN federation let researchers use their home credentials to access overseas datasets, reducing friction for collaborators. For sensitive datasets, data clean rooms and secure analysis platforms provide compute access without exposing raw records, an approach that has become standard for household survey data from the Pacific region.

Indigenous data sovereignty and Australian realities

Australia's international partnerships increasingly intersect with Indigenous data sovereignty, a field where communities assert the right to govern data drawn from their lands, waters, and traditional knowledge. The Maiam nayri Wingara Indigenous Data Sovereignty Collective and the Australian Indigenous Governance Institute have produced principles that align with the global CARE framework, but they also speak specifically to Aboriginal and Torres Strait Islander contexts, from savanna burning programmes in Arnhem Land to Indigenous ranger networks monitoring threatened species around Kakadu.

When agreements cross borders, these principles do not disappear. A project on Indigenous food systems that links Australian communities with Māori researchers in New Zealand or First Nations partners in Canada will need clauses that recognise collective ownership, benefit-sharing, and the right of communities to withdraw data if circumstances change. Major programmes such as the Global Environment Facility now expect Indigenous data governance to be addressed explicitly, so researchers who overlook it risk losing funding rather than gaining flexibility.

Practical choices follow. Data may be held in community-controlled repositories with delegated access rather than in a university server. Traditional knowledge labels, inspired by the Local Contexts initiative, can be attached to datasets to signal cultural protocols. Publications may need community approval before submission, and outputs may be required to return to Country in accessible formats. None of these arrangements need to slow a project down, but each needs to be written into the agreement rather than improvised later.

Negotiating an agreement that actually works

The final stretch of any data sharing negotiation is usually where momentum stalls. Lawyers tighten language, scientists lose patience, and months can pass while a single indemnity clause is redrafted. A few habits keep the process moving without sacrificing rigour.

Key elements worth insisting on:

• A clear data dictionary that names every dataset, its custodian, and its licensing terms
• Defined roles for a data steward, a technical lead, and an ethical review contact in each partner organisation
• A dispute resolution pathway that names a neutral forum, such as arbitration under the Singapore International Arbitration Centre rules
• An exit clause specifying what happens to data if a partner withdraws, including retention periods and continued access for published studies
• A review schedule, often every two to three years, to revisit terms as technologies and regulations evolve

Common pitfalls to avoid:

• Treating the agreement as a one-off legal hurdle rather than a living document that the partnership will revisit
• Borrowing clauses from unrelated contexts, such as clinical trials or commercial software licences, where assumptions about risk diverge sharply
• Leaving metadata standards unspecified, which later forces expensive reformatting
• Overlooking the need for plain-language summaries that non-legal readers in regional offices can actually understand
• Forgetting to register the agreement with internal research integrity or ethics offices, which can delay publication clearance

When these choices are made deliberately, the agreement becomes an enabler rather than a barrier. Researchers spend less time worrying about whether they are allowed to use a colleague's dataset and more time interpreting the patterns that those datasets reveal. For Australian science in particular, where partnerships span the Indo-Pacific and increasingly involve First Nations co-researchers, well-crafted agreements are the quiet infrastructure that keeps the work honest, shared, and ready to respond to the next food security shock.