AI & Global Food Security: A Focus on Precision Agriculture

AI & Global Food Security: A Focus on Precision Agriculture

Photo: -/AFP via Getty Images

[Link]

Q3: What are the limitations to and risks of AI-enabled applications to precision agriculture?

A3: While AI confers tremendous potential to improve precision agriculture and expand its adoption, there remain significant barriers that, if left unaddressed, could stymie this potential. The limitations and risks of AI-enabled technologies for precision agriculture mirror many of the challenges shared across the varied applications of AI.

Model bias is one of the most prominent risks, as AI models can reflect the biases of those that develop them, and models can be further biased by the datasets used to train them. The inherent complexity underpinning AI models and lack of model transparency can make it difficult for developers to easily address challenges with AI systems, and for potential users to establish trust in those systems. Various subsets of AI, including generative AI and machine learning techniques, also require access to potentially sensitive data, posing serious questions over data privacy, ownership, and stewardship. Negative outcomes from poor AI model performance (e.g. hallucinations, inaccuracies, and biased outcomes) raise the additional question of accountability for those potentially harmful outcomes, particularly as more individuals rely on AI models to inform important decisionmaking. Finally, there remain vulnerabilities for model manipulation by bad actors seeking to produce misinformation and disinformation that could result in both short- and long-term harms for users.

For farmers, these AI-related risks can manifest in general mistrust that can inhibit the acceptance and use of potentially powerful AI-enabled tools for precision agriculture. For smallholder and low-income farmers in remote or rural contexts, incorporating AI-enabled tools for precision agriculture carries additional limitations, namely, accessibility and usability of new technologies. As AI-enabled technologies continue to rapidly develop, the pace and success of diffusion among such farmers remains a present and future challenge. Factors that could inhibit scaling include low digital literacy, high upfront costs of implementation, local gaps in infrastructure (e.g., electrification), and varying rates of phone/digital ownership. If these barriers are not successfully overcome, low uptake by farmers could contribute to AI model inequity and decreased viability of AI-enabled tools for those individuals and communities who could most benefit.

Q4: How can AI-enabled precision agriculture improve food security?

A4: Twenty-eight percent of the global population, or 2.3 billion people, were moderately or severely food insecure in 2024. This is a slight decrease from a peak of 28.9 percent in 2021, but an increase of 6.6 percent in global food insecurity relative to just a decade prior. With the world's population projected to approach 10 billion by 2050, ensuring food security for all will require agriculture systems that produce a greater quantity of more nutritious food in more sustainable ways. The ability for precision agriculture to contribute to these goals will be dependent, in part, on the ability of AI to increase the accessibility and utility of precision agriculture itself.

In addition to the examples discussed above, the variety of AI technologies that could be applied to agriculture-including, for example, computer vision and robotics-offer further benefits for smallholder farmers. Importantly, applying AI to precision agriculture can make otherwise expensive tools more affordable. For farmers facing climate shocks and stresses, fluctuating markets, and assistance program bureaucracy, high upfront costs often limit uptake of potentially beneficial agricultural practices that reduce expenses, mitigate environmental impacts, and increase farm productivity and profitability. Collectively, the various types of emerging AI systems could be used to reinforce one another to more effectively monitor crop productivity, measure environmental conditions in real-time, and forecast local conditions, ultimately providing a more comprehensive toolkit for enhanced decisionmaking.

Increasing agricultural productivity could introduce more food into global markets, helping to buffer markets against shocks, reduce food costs, and increase agricultural profits. As smallholder farmers and farm laborers are typically net food buyers, increased revenues, alongside lower food prices, would likely increase their access to healthy diets. Lower food prices that result from increased quantity and stability of yields and decreased input costs could also benefit cities and towns, where most of the world's food-insecure people live.

As technologies evolve, however, policymakers should recall that "simply growing more food globally will not lead to more food-secure countries," as the U.S. intelligence community cautioned in 2015. Improving agricultural productivity and on-farm operations can have numerous benefits for agricultural markets and farming communities, while many other conditions must be met to improve food security in diverse settings worldwide.

Q5: What U.S. and global policies could promote the development and application of AI technologies for precision agriculture while minimizing risks?

A5: The rapid integration of AI across agricultural production portends what many consider to be revolutionary potential. Broader access to AI-enabled technologies that result in better decisionmaking and improved precision agriculture practices could play a significant role in sustainable intensification, conferring higher yields with lower ecological impacts. Should these gains in food production be met with advances across the agrifood system-such as hunger forecasting, crop breeding, supply chain optimization, and urban food infrastructure-then AI could facilitate positive outcomes for food-insecure populations and farmers while making progress toward global environmental and economic benchmarks.

These potential far-reaching benefits of AI-integrated global food systems are already on the minds of policymakers around the world, including in the United States, the African Union, and the European Union. In this context, it is important to underscore the responsibility of national and international governments in creating the conditions in which AI-enabled technologies can have maximum benefits for agriculture and food security. Such conditions include

• defined standards for data sharing and privacy that protect farmers' data but allow for AI model developers' access to high quality, robust, and locally specific datasets;
• agreement to avoid duplicative efforts and work towards delivering interoperable applications;
• adoption of a solution-agnostic approach, prioritizing the achievement of positive results for farmers and their communities over the promulgation of any specific tool; and
• incentivizing data sharing by delivering effective and equitable solutions to farmers.

Without the trust of farmers, secure and open data-sharing norms, and an environment that prioritizes actionable solutions, it is possible for AI to compound risks among already vulnerable populations that stand to gain the most from agricultural innovations. Anticipating and proactively responding to the known risks associated with AI-enabled agricultural technologies will be essential to avoiding the negative consequences of previous agricultural innovations that have led to modern agriculture's outsized environmental impact.

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