President Trump recently signed Executive Order 14415, “Securing America’s Defense Supply Chains and Ensuring Domestic Acquisition of Critical Materials”, which states that “the United States must secure its supply chains against physical, cyber, and economic subversion”, and that finished equipment and “the critical materials and components necessary to manufacture, maintain, sustain, and repair that equipment, are sourced domestically or from allied nations.” Among other provisions, the Executive Order states that the Department of War will, with the help of the defense industrial base, engage in supply chain mapping and illumination that will “map national security vulnerabilities as they relate to the sourcing of key raw materials or other links in the supply chain, using any tools and technologies to include artificial intelligence to assist in doing so.”

The goal to reduce risk in defense supply chains is extremely worthwhile. A strong national defense requires secure and resilient supply chains. Defense supply chains are responsible for supplying the flow of critical goods, services, and technologies that support national security priorities at home and abroad, and vulnerabilities in the supply chain imperil the success of these missions. Not only can the flow of materials be delayed and disrupted, supply chain cyber espionage can expose intellectual property and sensitive information to adversaries. Even worse, examples like the exploding pagers in Lebanon show that supply chains can be weaponized against the very people who rely upon them.

The use of AI to map supply chains presents both opportunities and challenges. AI excels at crunching large volumes of messy, unstructured data, which is the type that tends to be associated with complex, global supply chains. AI is already being applied across the supply chain for other use cases like supply and demand forecasting, warehouse and transportation management, and inventory optimization. For supply chain mapping, while visibility into first and second tier suppliers may be feasible, many defense programs have five, six, or more tiers of suppliers, making it difficult to manually map out the entire supply chain from raw inputs to finished products. AI is a perfect candidate to fill this gap.

However, like any model, the output quality depends on the quality of the inputs, highlighting the importance of data integrity. Additionally, AI has the tendency to hallucinate, and the hallucinated results are often presented confidently. These data integrity issues, coupled with the complexity of numerous tiers of suppliers and the potential for AI hallucinations, require careful review and validation of supply chain mapping results to ensure that the insights are credible and actionable.

Supply chain mapping, if done correctly, is an important step in highlighting vulnerabilities that can be prioritized and mitigated. Shedding light on defense supply chains can identify harmful foreign dependence on inputs like critical minerals and electronics that are essential to national security. While AI can be leveraged for its computational capabilities, it also has the tendency to hallucinate. In the context of an illumination exercise, this could mean illuminating a risk that isn’t there, or worse, not illuminating a risk that is there. As with paper maps and GPS, the supply chain map’s accuracy is essential for reaching the destination – in this case, secure and trusted supply chains

Zachary A. Collier is assistant professor of management and director of the Center for Applied Analytics at Radford University, and a visiting scholar at the Center for Hardware and Embedded Systems Security and Trust (CHEST).