Open source in agrifood — cross-cutting framework: Mozilla findings, agricultural cultural ethos, and the field guide's existing threads

Global

Content

This unit synthesises how open source weaves through the existing field guide’s 49 units, anchored on the Mozilla State of Open Source AI report (July 14, 2026) and the A Growing Culture “Open Source Ethos” framing (January 25, 2023).

Open source is not a foreign technical import for agriculture — it’s a contemporary expression of an agricultural heritage that has been open for millennia. The Mozilla report provides the substantive data anchor; the A Growing Culture framing provides the conceptual bridge.

Mozilla 2026 headline findings (substantive anchors)

• 3% performance gap with proprietary models — open source is no longer playing catch-up.
• 50x cost reduction in three years.
• ~33% of real-world AI usage runs on open models.
• Only 4% revenue capture — value is real; revenue isn’t flowing back to open ecosystem.
• China/East Asia at 89% adoption — far ahead of West.
• 47 countries restricting foreign processing for critical workloads.
• 79% of developers use open models, only 51% deployed in production.
• “The real fight has moved beyond the model” — agentic harness matters more.
• 93% default-approval of AI agent requests (“consent fatigue”).

Mozilla’s framing — applied to the field guide

Mozilla CTO Raffi Krikorian: “Open source AI has reached a turning point. It’s no longer about expanding access to models; it’s about who has the power to shape, audit, and improve them.”

Applied to agrifood: The work ahead isn’t about which model is best for agriculture. It’s about who controls the deployment harness — the data, the data-rights framework, the consent management, the agent infrastructure. This maps directly to the field guide’s data-governance and data-rights-framework work.

Five Mozilla findings mapped to the field guide

The conceptual bridge — A Growing Culture’s “Open Source Ethos”

The A Growing Culture Substack (January 25, 2023) explicitly bridges software open source, scholarly open access, biological open-pollination, and traditional agricultural seed-saving:

“Manifestations of open are seen in our agricultural heritage, owned by all and fabricated over generations. The boundless mixing of genetic materials seen in open pollination boosts the overall vigour of plants and regenerates the abundance of nature. There is also a long history of open in many traditional cultures around the world. Local knowledge, often collectively inherited, shared orally, and passed over jurisdictional borders and geographies, is the backbone of traditional agriculture and biodiversity preservation as farmers save and trade seeds.”

The conceptual move: Open source in agriculture has multiple genealogies — software, scholarly, biological, cultural. The agricultural / biological lineage is older than the software lineage. The farmer-cooperative / commons cultural tradition (Pillaud’s framing) is the European-explicit expression of this lineage.

How open source already weaves through the field guide

Layer 1 — Open-source standards (data interoperability):

• OADA — RAML API specification on GitHub. Distributed model.
• JoinData — operational non-profit data cooperative.
• GAIA — open-source core of agricultural data; generative AI agents.

Layer 2 — Open data frameworks:

• CARE Principles for Indigenous Data Governance — 1,808 citations, multilingual translations.
• FAIR Principles — scientific data.
• GODAN 2.0 — Global Open Data for Agriculture and Nutrition.
• CGIAR Platform for Big Data in Agriculture — open data, Responsible Data Guidelines.
• Open Data Ecosystem unit captures this layer.

Layer 3 — Open cooperative / commons models:

• JoinData (Netherlands) — first agricultural data cooperative.
• AGUAPAN (Peru) — custodian-farmer network; cooperative cultural fit.
• NAPDC / Ag Data Coop (US) — federally-funded cooperative framework.
• La Ferme Digitale / GAIA (France) — agtech association; cooperative / commons.
• EMILI Innovation Farms — industry-led non-profit.
• AIVA Network — multi-actor coalition validation infrastructure.

Layer 4 — Open Indigenous Data Governance:

• CARE Principles — already noted.
• IEEE 2890-2025 — first global standard for Indigenous data provenance.
• Indigenous Navigator (IWGIA) — open methodology.
• WIPO Treaty on Genetic Resources (2024) — open multilateral framework.

Layer 5 — Open-source implementation examples in agriculture:

• FarmOS — open-source farm record-keeping.
• FarmVibes.AI — Microsoft Research’s open-source geospatial ML.
• eLocust3 (FAO) — open-source locust tracking.
• GAIA — already noted.
• OpenAgri (EU) — industry-driven open-source solutions.

Open-source challenges named in the field guide

• Mozilla 2026: Only 51% of open models deployed in production. Infrastructure is the gap, not quality.
• OpenAgri (April 2024): “A large proportion of OS/free projects have not continued to have an active community and tend to fizzle out over time. Projects developed with limited funding often struggle to maintain momentum and attract new contributors.”
• Mozilla 2026: 4% revenue capture — economic sustainability is unsolved.
• Pillaud’s framing: “Unicorns do not exist in agriculture. The French agri-tech industry is not doing well, and funding from private investors remains insufficient. Using open source and developing digital commons are the obvious solution.”

The institutional layer — UN / FAO / Digital Public Goods

The Vincent Martin (FAO Director of Plant Production and Protection) framing from the OSPOs for Good conference:

“Open-source technological solutions are revolutionizing agrifood systems not simply as a technological phenomenon. It democratizes access to innovative tools and fosters collaborative approaches to address global challenges. It is an excellent example of Open innovation, a collaborative approach that brings together external ideas, technologies, and resources to solve problems more effectively. It is a way of working to create better solutions that enable farmers, researchers, and policymakers to freely adapt through a more inclusive and equitable approach to agricultural innovation.”

The UN Secretary-General has identified open-source solutions as pivotal in the context of UN 2.0 and the quintet of change for SDG achievement. The Digital Public Goods Alliance + FAO certification process produces certified open-source tools as “digital public goods” — the institutional expression of the open-source movement applied to development.

What this unit is doing in the taxonomy

Cross-cutting framework unit that synthesises how open source already weaves through the existing field guide. Anchored in (a) Mozilla 2026 substantive data, (b) A Growing Culture conceptual bridge, (c) FAO/UN institutional layer, (d) the field guide’s existing units.

Distinct from:

• Mozilla State of Open Source AI report unit (mozilla-state-of-open-source-ai-2026.md) — the substantive data anchor; this unit is the agrifood-specific framework synthesis.
• Open data ecosystem unit (open-data-ecosystem.md) — agricultural open data; this unit is open-source AI in agrifood.
• Data cooperatives / commons architecture unit (data-commons-architecture.md) — operational/framework/standards layers of data commons; this unit is broader (software + data + cooperative + Indigenous).

Why it matters for talks

This unit lets a talk:

1. Anchor on Mozilla 2026 for substantive figures (3% / 4% / 89% / 51% / 93%).
2. Bridge to agricultural heritage through A Growing Culture framing.
3. Surface the field guide’s existing threads — JoinData, OADA, GAIA, NAPDC, CARE Principles, Indigenous Navigator.
4. Connect to UN/FAO institutional layer for international policy engagement.
5. Name the substantive challenges — infrastructure gap (Mozilla), sustainability (Pillaud), deployment rate (OpenAgri).

Critical context

• Open-source AI in agrifood is technically feasible (Mozilla 3% gap) but deployment-sparse (51% rate). The work ahead is infrastructure, not capability.
• Open-source AI captures 4% of revenue (Mozilla) — the funding model is structurally different from proprietary AI in other sectors. This validates Pillaud’s “unicorns do not exist in agriculture” framing.
• The cooperative / commons response is not a workaround but the substantive alternative for sustainable agrifood AI.
• The agentic harness (the deployment layer) is the governance layer. The data-governance / data-rights-framework work in the field guide is about this layer, not the model layer.
• China and East Asia’s 89% open-source AI adoption has direct implications for Chinese agritech AI deployment (DJI, XAG, Alibaba, Pinduoduo all have open-source / open-data components).