By: Jeffrey D. Bean
U.S. president Donald Trump rolled out his AI Action Plan on July 23, in conjunction with a trio of executive orders to operationalize aspects of the plan, involving exporting AI infrastructure, building out data centers, and removing the “bias” of past diversity initiatives. With an overriding emphasis toward recommendations to lowering barriers to deployment and innovation, the plan intends to jolt U.S. government and industry into action on artificial intelligence. While largely domestically orientated, the plan has significant implications for U.S. allies and partners, both in terms of commercial engagement and national security considerations.
The AI Action Plan is organized under three pillars: innovation, infrastructure, and international diplomacy and security. It argues for urgency in retaining U.S. leadership in AI and is broad in its scope and responsibilities. While involving nearly 30 U.S. agencies, the heaviest potential burden is on the Department of Commerce. It also recommends the creation of a Chief Artificial Intelligence Officer Council to coordinate interagency collaboration on AI adoption.
The international implications are important. The plan directs the United States to provide technology offerings (consistent with export controls) to allies and partners, including full-stack AI technology packages for export. That could include AI-optimized computer hardware (semiconductors, servers, and accelerators), data center storage, cloud services, and networking; data pipelines; AI models and systems; cybersecurity measures; and AI applications for specific use cases (such as education or agriculture). This effort at competing with other hardware stack providers (e.g. China) is likely to be welcomed both by the U.S. technology industry and partner countries. The new AI plan also aims to ensure U.S. government leads in evaluating national security risks of frontier AI models and to counter Chinese influence on tech in international governance bodies.
But the plan also emphasizes protection measures, including closing export control loopholes and compelling partners to align with U.S. export control regimes, intellectual property protection, and geofencing for chips. Challenges include resources for enforcement, both within the Bureau of Industry and Security (BIS) at Commerce and a proposed new role for the intelligence community on combating chip diversion. There may be associated costs to chip performance, production timelines, and market efficacy that make geofencing unviable while also failing to address advanced compute access via cloud services from Chinese, Russian, or other entities.
A few other concerns to implementation are worth considering. Attempts at challenging Chinese leadership at multilateral bodies will be hindered by recent cuts and firings at the State Department, including in the Office of the Special Envoy for Critical and Emerging Technologies and the Bureau of Cybersecurity and Digital Policy. Many aspects of the plan depend on a complex network of global technology supply chains and foreign talent to design and operationalize the stack, so it cannot be an America only effort. Potential partner countries may also be more reluctant to engage with the United States given the looming hurdle of trade negotiations. The plan also does not address important shortfalls in U.S. funding for research and development or in basic science, technology, engineering, and mathematics (STEM) capabilities. There is no clear indication that international research cooperation will expand or that tensions between the federal government and states will be resolved. Regardless, allies and partners of the United States now have clear basis for a shift in the articulation of U.S. approach to AI and technology sales and the opportunity to engage in that effort.
Jeffrey D. Bean is the Program Manager for Technology Policy and Editor at ORF America.