When your printer orders its own toner, you probably don't think about national security, international trade policy, or the global supply chains that make artificial intelligence possible.
But perhaps you should.
In a recent episode of Voices of Biocomplexity, University of Virginia Biocomplexity Institute researcher Vidya Mani explains how the rapidly expanding use of artificial intelligence is changing our understanding of supply chains, and why the same principles that help businesses manage physical goods can help us identify vulnerabilities in AI systems.
The conversation begins with a deceptively simple question: What does artificial intelligence have to do with supply chains?
The answer, Mani explains, is almost everything.
Beyond Trucks and Shipping Containers
For most people, supply chains conjure images of cargo ships, warehouses, delivery trucks, and products moving from factories to consumers. But these physical movements represent only part of a much larger system.
Consider ordering groceries online. Behind the seemingly simple transaction are software systems that process orders, determine availability, recommend substitutions, and coordinate deliveries. Increasingly, AI-powered agents can perform some of these tasks automatically.
AI has become part of the supply chain itself.
Yet there is another dimension to this relationship: AI systems depend on supply chains of their own. Before an AI model can answer a question, analyze information, or complete a task, it needs computer chips, software, training data, computing infrastructure, cloud services, and people with specialized expertise.
Each component introduces dependencies. And each dependency can create vulnerabilities. If a critical chip becomes unavailable, a cloud service experiences an interruption, or a software component is compromised, the effects can extend far beyond the original point of failure.
Familiar Risks in an Unfamiliar World
Mani argues that many of the risks associated with AI are not entirely new. They resemble challenges that supply chain researchers have studied for decades.
A manufacturer that depends on a single supplier for an essential component faces concentration risk. An AI developer relying on a limited number of computing providers or foundational models may face a similar vulnerability.
Physical products can be affected by transportation bottlenecks, manufacturing disruptions, or geopolitical restrictions. AI systems can be affected by limitations in access to hardware, software, data, and computing resources. But there is an important difference.
Physical goods are commonly classified using standardized international trade codes, making it possible to track products and understand where they originate.
Software and AI models do not fit neatly into those same systems of classification. Their components, origins, and dependencies can be considerably harder to identify. This lack of visibility makes assessing risk more complicated.
Mani also distinguishes between open-weight and closed-weight AI models, explaining how different approaches to making model parameters available can affect the ability of organizations to examine, adapt, and deploy AI technologies. For organizations adopting AI, understanding these differences is becoming an increasingly important part of managing risk.
When Technology Becomes National Security
The conversation also explores why seemingly unrelated headlines about AI companies, semiconductor restrictions, tariffs, and international competition are connected. Mani describes a fundamental tension among three priorities: economic utility, public privacy, and national security.
Widespread access to technology can encourage innovation, competition, and lower costs. But governments and organizations must also consider how sensitive information is protected and whether strategically important technologies could be used in ways that threaten national security. These competing priorities help explain why governments establish export controls, restrict access to certain technologies, and invest in domestic technological capabilities.
The consequences extend across global supply chains. Restrictions intended to protect one country's technological advantages can influence where companies obtain computing resources, which AI models organizations adopt, and how businesses design their systems.
Understanding AI, therefore, requires understanding the economic and political systems in which it operates.
Building Resilience into AI Systems
Despite these challenges, Mani emphasizes that organizations are not powerless. In fact, some of the most useful lessons come from traditional supply chain management.
Rather than simply adopting an AI system because it is available, organizations should begin by identifying what they actually need the technology to accomplish.
They should understand the components on which the system depends, evaluate potential vulnerabilities, establish appropriate safeguards, and test how the system performs under different conditions. This requires more than technical expertise. It also requires knowledge of the organization's operations, objectives, and risks.
Mani cautions against becoming so dependent on convenient, ready-made technologies that organizations lose sight of the systems supporting them. Instead, she encourages organizations to design AI capabilities around their specific needs, with attention to both performance and resilience.
Seeing the Connections
Perhaps the most important lesson is that supply chains are not defined by trucks, ships, or warehouses. They are defined by relationships and exchanges. Those relationships can involve physical products, digital services, computing resources, or information.
And as AI becomes increasingly embedded in everyday activities, understanding those connections becomes essential.
The same systems thinking that helps researchers understand disruptions in global trade can help organizations evaluate the opportunities and vulnerabilities of artificial intelligence.
The question is no longer simply what AI can do.
It is also what AI depends on, who controls those dependencies, and how we can build systems that remain reliable when conditions change.
Listen to or watch The Supply Chain I Couldn't See, featuring Vidya Mani, on the latest episode of Voices of Biocomplexity.