AI and Global Food Security: A Focus on Crop Breeding

Critical Questions by Zane Swanson, Emma Curtis, and Caitlin Welsh

Published March 26, 2026

In January 2026, the CSIS Global Food and Water Security Program convened the third in a series of roundtable discussions to better understand the emerging benefits-and potential risks-of artificial intelligence (AI) applications for global food security. The first two roundtables, which took place in July and December 2025, focused on AI applications for precision agriculture and food security early warning systems, respectively. The January convening, which had a specific focus on AI-enabled applications to crop breeding, brought together representatives from a range of companies, research institutes, nonprofit organizations, international organizations, and U.S. government agencies. The following questions were explored, illuminating the state of AI integration into crop breeding today and providing an outlook on its potential in the future.

Q1: What is crop breeding?

A1: Crop breeding is the foundation of the modern global agrifood system, emerging more than 10,000 years ago with the first instances of the domestication of wild plants. The intentional selection of beneficial traits in plants and seeds by farmers and breeders (i.e., artificial selection) introduced the broad diversity of domesticated crops that make up the human diet. Across history, the continued development of improved crops has been integral to all major agricultural advancements. In the last century, scientific progress has driven particularly remarkable agricultural productivity.

Key Terms

Modern crop breeding is defined by the application of scientific techniques to improve the genetic potential of plants used for food, fuel, manufacturing, ecosystem services, and other human activities. This is achieved through the intentional manipulation of plant genetics, either through "conventional" processes or genetic engineering (GE), to develop crop varieties that are better adapted to the environments and management practices in which they are grown, thereby increasing production quality and yield.

Q2: Why is crop breeding important to agriculture today?

A2: The modern agrifood system-and the crops that support it-face myriad, intensifying pressures. Physical and biological threats force significant crop losses, with pests and pathogens accounting for harvest losses that range as high as 40 percent annually. At the same time, changing precipitation patterns, rising temperatures, and the increasing incidence of extreme weather events can cause tens of billions of dollars in lost and damaged harvests. Breeding novel crop varieties that can endure these pressures is essential for enhancing present and future food security, as well as bolstering the economic security of farmers and those employed across the agrifood system-some 22.1 million people in the United States and about 1.23 billion people globally.

Like those across the history of agriculture, crop breeders today are focused on developing improved varieties that are more resistant to biological threats, more resilient to environmental hazards, and more productive and nutritious, all while requiring fewer inputs. Unique to the present moment, however, is the pressure on the world's farmers to feed a growing global population while facing these ecological stressors. By 2050, food demand could increase by as much as 56 percent, with crop production demand growing by at least that much. In the absence of accessible, improved crop varieties, it will become increasingly difficult for producers-smallholder farmers and large agrifood companies alike-to keep pace with agriculture's evolving risk landscape. As such, the food, nutrition, and economic security of millions could suffer for it. Already, 2.6 billion people are unable to afford the least expensive healthy diet. Today, high food prices increase consumers' food budgets, forcing many to purchase less nutritious food or to simply consume less food overall.

Today, however, there are two AI-enabled revolutions changing the way that crop breeding is conducted-one biological and one digital. The confluence of these revolutions will undoubtedly play a role in changing the agricultural landscape for both producers and consumers, but how they combine to change the fate of food-insecure populations remains to be seen.

Q3: How can AI improve crop breeding?

A3: AI-enabled technologies offer the potential to improve both the process and outcome of plant breeding.

Process: Crop breeding is data-, time-, and labor-intensive. It often relies on researchers creating breeding strategies using manual field trials and phenotyping techniques, while simultaneously considering the complex interactions between plant genotypes, environmental variables, and management strategies (G x E x M) that can produce a range of unpredictable outcomes. Even with improved tools that help breeders better understand the genetic underpinnings of phenotypic variation, the development cycle for the production of a new crop variety can take up to a decade or more. However, AI-enabled crop breeding technologies promise to dramatically shorten this timeline, accelerating work across most, if not all, of the crop breeding process. This promise is shaped by three AI-powered advantages: improved data management, enhanced data collection and analysis, and novel genomic tools.

1. Data Management: AI is already playing a role in the development of more effective data systems for plant genetic materials, enabling access by more people to more data. The genetic data stored in an individual plant specimen is enormous-the wheat genome, for example, contains about five times more DNA than the human genome. Today, decades of genetic and phenotypic data are stored across formats that are either incompatible or inaccessible for researchers. The world's genebanks, which collectively store millions of plant samples, represent a vast sum of yet-to-be discovered data, so-called dark data. These dark data are a reservoir of heretofore unrealized crop genetic potential. The International Rice Research Institute (IRRI), for example, stores more than 130,000 rice samples, yet has only been able to use about five percent of its collections for breeding efforts to date. AI presents a unique opportunity to both discover these data and harmonize them as an interoperable resource. To shed light on these data, organizations like the Crop Trust are turning to AI to accelerate the digitization of their vast genebank collections to better protect the biodiversity they already store and make it more accessible to the world's crop breeders.
2. Data Collection and Analysis: The traditional manual processes of phenotypic data collection and analysis for crop breeding are hugely labor- and time-intensive and often rely on limited and variable expertise. There are several AI-enabled systems that are individually-and collectively-set to enhance the speed, accuracy, and consistency of decisionmaking for crop breeding. Initiatives like the CGIAR's Artemis Project are leveraging computer vision technologies for the rapid identification of plant traits that can then be read into machine-learning models for more informed decisionmaking, even in low-resource settings. When paired with AI-enabled robotics technologies, scientists are also quickly accelerating insights into the gene-trait interactions that are central to developing selection models for improved crop varieties. Additionally, feedback loops between farmers and crop breeders, enabled by the increased adoption of digital and AI-enabled tools, can improve both guidance for the use of novel seed varieties by farmers and information on real-world performance for breeders.
3. Genomic Tools: AI-enabled technologies are fundamentally changing the application and accessibility of genetic engineering techniques for crop breeding. Genome sequencing-accelerated and transformed by AI tools-can turn huge quantities of stored genetic diversity into actionable information for breeders, enabling the accurate prediction of desirable traits before genes are even introduced into plants. A trait that once took a generation or more to stabilize in a crop can now be introduced by targeted intervention and then rapidly validated through field testing. When these predictive genomics tools are combined with AI systems like Google DeepMind's AlphaFold, scientists and crop breeders can go even one step further, modeling plant-pathogen interactions to discover new disease-resistant genes for crops at an unprecedented pace. At the same time, the AI tools are likely to significantly lower the barriers for genomic research. Organizations like GetGenome are leveraging AI tools to expand access to genomic techniques that can aid in generating improved crop breeding efforts across the Global South.
   

Outcome: AI functions as a force multiplier across the entire crop breeding process, from data management, collection, and analysis to the application of these data through conventional and GE techniques. By improving the pace, precision, and scope of crop improvement, AI-enabled breeding tools help provide farmers with seeds that are better adapted to current and near-term ecological conditions.

These tools support the development of crop varieties with enhanced resistance to biological threats (e.g., pests and diseases), strengthened resilience to environmental stressors (e.g., drought, flood, and extreme heat), reduced input requirements (e.g., water, fertilizer, and pesticides), and improved nutritional profiles. Critically, the expansion of AI-enabled tools and crop data discovery can also broaden the diversity of improved varieties available to farmers, thus expanding planting options that strengthen the economic resilience of farmers and the ecological resilience of their agricultural lands.

Q4: What are the limitations and risks of AI-enabled crop breeding techniques?

A4: AI is likely to catalyze major transformations in crop breeding. However, the speed of these transformations will depend on stakeholders' abilities to overcome significant barriers to access AI technologies and policymakers' abilities to effectively regulate them.

While AI tools could significantly expand access to cutting-edge techniques and novel data for crop breeding, most advanced crop breeding work is either heavily concentrated in the Global North or is led by Western companies and scientists in the Global South. Such approaches risk excluding those counties and innovators for whom AI-enabled advances in crop breeding hold the greatest relative gain. Similarly, while AI has lowered the barrier for entry for many individuals and startups, unequal access to genetic data-particularly compared to larger agrifood companies-remains a significant obstacle to innovation. Should AI-enabled crop breeding techniques or data be unequally deployed, there is significant risk that these technologies could exacerbate existing agricultural inequities that actually hinder the achievement of global food security.

AI tools for crop breeding also face limitations across the regulatory environment. As models for AI governance begin to emerge in leading countries, it will become increasingly important to align AI regulatory frameworks with existing national regulations that affect crop breeding, such as policies concerning genetically modified crops, biotechnologies, and agriculture. Alignment across international regulatory frameworks will likely prove particularly challenging, too, as current geopolitical uncertainty and tensions over data and technology sharing grow. Such uncertainty is likely to serve those who already have the economic capacity to overcome a complex, expensive regulatory environment. Should these conditions emerge, the majority of those who would stand to benefit most from the products of AI-enabled crop breeding are the least likely to obtain fair and equitable access to those products.

Q5: What U.S. and global policies are needed to advance AI-enabled technologies for crop breeding?

A5: Innovative, AI-enabled crop breeding strategies are vital to the development of more resilient, higher-yielding, and more nutritious crops produced with fewer inputs and lower impacts on local ecologies. Modern breeding techniques applied during the latter half of the twentieth century drove significant yield growth for some of the major food crops-more than 200 percent for wheat, 150 percent for maize, and 100 percent for rice. Despite this, agricultural yield gaps remain across much of the Global South, where food insecurity and malnutrition are also more prevalent-Africa and Asia account for about 630 million of the 673 million people (93.6 percent) who were undernourished in 2024.

Today, the pace of technological advancement and the pace of ecological change are both moving faster than decisionmakers can manage. Creating an enabling environment for emerging crop breeding technologies that benefit researchers, private sector actors, farmers, and the public alike requires policies that focus on developing AI-technologies and that look ahead to their applications across various sectors, including agriculture and crop development. To achieve this, policymakers can

• prioritize support for foundational sciences that serve as the bedrock for innovations in applied research for AI, biology, and agronomy, which are all essential to the advancement of crop breeding;
• support institutional capacity building, including for genebanks, improving the speed of data discovery and the deployment of innovation to farms;
• proactively develop coherent regulatory pathways and robust biosafety standards to avoid mismatched policy frameworks and overregulation that could stymie the development of AI-enabled tools for agricultural biotechnologies before they can be deployed; and
• treat AI, genomics data, and advanced crop breeding technologies as tools for collaboration and capacity building, particularly with partners in the Global South, to realize long-term gains for food systems.
  

The challenges facing present and future agricultural production will not be overcome by any single solution. Rather, they will require a combination of many distinct and interconnected solutions, such as improved early warning systems, emergent precision agricultural technologies, innovative guidance for farmers, and revolutionary crop breeding techniques. Among these solutions, the impact that the effective deployment of AI-enhanced crop breeding could have on agricultural production demands the focus of decisionmakers who are interested in both the development of the global agrifood industry and the improvement of global food security.

Zane Swanson is the deputy director of the Global Food and Water Security Program at the Center for Strategic and International Studies (CSIS) in Washington, D.C. Emma Curtis is a research associate with the Global Food and Water Security Program at CSIS. Caitlin Welsh is the director of the Global Food and Water Security Program at CSIS.

Programs & Projects

Critical Questions by Emma Curtis, Zane Swanson, and Caitlin Welsh - August 13, 2025

Critical Questions by Emma Curtis, Zane Swanson, and Caitlin Welsh - January 20, 2026