Sen. Jim Banks (R-IN), an active player on AI policy as a Senate Banking and Armed Services member, is urging White House economic advisor Christopher Phelan to examine the benefits of open-source AI models and address potential incentives to bolster the open-source ecosystem, while also touting the role of existing U.S. semiconductor export controls.
“I am writing about the importance of American leadership in both frontier artificial intelligence (AI) development and the global diffusion of AI models, including open-source models,” Banks says in the Aug. 13 letter to Phelan, chair of the president’s Council of Economic Advisers.
“America faces two related but distinct AI competitions. The first is a vertical race to the frontier to ensure that America, not China, develops the world’s most capable AI systems. The second is a horizontal diffusion race to ensure American, not Chinese, models and technical ecosystems become the default for businesses, researchers, and governments around the world,” Banks writes.
He says, “America cannot afford to see Chinese open models proliferate and burrow into the global economy only to be weaponized, like rare earths, at a time and place of China’s choosing. We need an American AI ecosystem that wins both the vertical and horizontal races.”
Banks asks Phelan to look into three sets of questions and to schedule a “staff-level briefing” within 60 days:
- Economic incentives and open-source. Do private incentives systematically undervalue the economic and national-security benefits of American open models? What benefits accrue broadly to the United States but cannot be captured by individual companies? How does the availability of strong American open models affect innovation and competition among startups, small businesses, and independent developers?
- Tools for promoting U.S. open-source leadership. What policy tools or economic incentives could address potential market failures and strengthen the U.S. open-source ecosystem? Which interventions would most effectively support the development, distribution, and adoption of competitive American open models? How can such policies support startups, small businesses, and independent AI developers?
- Preventing China from cheating. What are the quantifiable economic benefits of semiconductor export controls for both the frontier race and the diffusion race? What is our best estimate of how much existing export controls have strengthened U.S. AI leadership in each race, and how much have these advantages contributed to gains in U.S. productivity, growth, investment, or global market share? What economic costs would result from failing to prevent large-scale distillation or enforce existing controls?
Banks says, “Open models can generate broad economic benefits by lowering barriers to entry, reducing development costs, strengthening competition, and enabling downstream innovation. Frontier developers invest large sums to develop models and frequently rely on controlled access to recover those investments.”
“For a given frontier model developer,” he writes in the letter, “releasing weights openly could reduce revenue, transfer capabilities to domestic competitors, or attract scrutiny if open models create harm. As a result, frontier developers may underproduce open models due to their structural incentives.”
Banks says, “The United States must think creatively about the types of incentives that may change this calculation and promote a strong American open-source ecosystem.”