America built the world’s most successful modern economy not simply by creating innovative technologies, but by spreading them throughout the economy. At the turn of the 19th century, electricity transformed productivity as factories reorganized around it. Nearly one hundred years later, the internet created even greater value when businesses used it to build new services and industries. Today, artificial intelligence poses the latest test to that formula.
Read more Education shuffling is worse than doing nothing
The United States has developed many of the world’s most capable closed-source AI models and designed advanced chips that supply extraordinary computing power. Yet it has not translated those strengths into an equally competitive open-source ecosystem that businesses can use to develop their own products. If it wants to realize the full potential of the technology, America must find a way to champion both model types.
This does not require choosing between closed and open-source AI. Closed-source models will remain essential to pushing the technological frontier and serving customers who want sophisticated, ready-to-use systems. Open models perform a complementary economic function: they can be downloaded, operated, and adapted by users, allowing more businesses to innovate and making the technology more accessible across the marketplace.
A startup can build a product without paying a provider for every query. A manufacturer can customize a model around its production process. A hospital can keep sensitive information on its own systems. Farmers can use AI to monitor crops, forecast yields, conserve water, and detect pests. Open access gives each greater freedom to experiment and solve problems the original developer may never have anticipated.
That is how a breakthrough becomes a source of economy-wide growth. The largest gains from a general-purpose technology often emerge from businesses that adapt it and discover uses its inventors did not envision. Linux Foundation research found that two-thirds of surveyed organizations considered open-source AI cheaper to deploy than proprietary alternatives. Lower costs allow startups and midsize businesses to test ideas, enter markets, and compete.
The economic case for American open-source AI would be compelling even if China were not part of the equation. But the rapid improvement of Chinese open models makes closing America’s gap more urgent. Chinese developers are competing for global market share by offering capable, inexpensive models that businesses can adapt to their own needs. Unless American alternatives keep pace, China could translate its advantages in cost and flexibility into a dominant position in the global open-model market sooner than many American leaders expected.
Moonshot AI’s Kimi K3 illustrates how quickly that competitive landscape has changed. For months, estimates placed Chinese models six to 12 months behind American frontier models, reinforcing confidence that U.S. closed-source leadership would endure. Kimi reached the top tier of global models, ranking ahead of Anthropic’s Claude Opus 4.8 in one broad composite evaluation while costing 40 percent less to run. It does not outperform the best American systems on every task, but it undermines the assumption that superior closed models alone will preserve America’s lead.
Read more Francesca walks the plank
Perhaps the largest danger is that American technological leadership will fail to translate into market leadership. Stanford researchers have documented growing adoption of Chinese open-weight models, including by American companies. As more businesses build on those models, developers create compatible tools, workers acquire model-specific skills, and investors finance complementary products. Those network effects can create a durable advantage that competitors will struggle to dislodge.
Concerns about the misuse of open models deserve a serious response. But regulation carries economic costs of its own. Carl Benedikt Frey, a professor at Oxford University, argues in the Wall Street Journal that China’s rapid AI advancements are thanks to Beijing’s effective yet risky approach of “move fast and regulate later” where the government oftentimes allows firms to scale in regulatory gray zones before later intervening. This is a much different dynamic than Washington’s current ruminations on whether to crack down on AI software.
Broad restrictions on American open-model development would raise barriers to entry, protect incumbent providers, and steer cost-conscious businesses toward foreign alternatives that remain available. Safeguards should therefore focus on specific credible threats, preserving competition and experimentation across the much larger market for beneficial uses.
The Open Secure AI Alliance shows how the market is already developing ways to manage those risks. NVIDIA, Microsoft, IBM, Cisco, Palantir, Hugging Face, and many other companies are collaborating on open tools that make AI systems easier to test, adapt, and secure. Shared tools can lower security and compliance costs across the market. The alliance demonstrates that companies with different business models can develop common safeguards without treating openness as a defect or suppressing competition.
Policymakers should preserve that competitive environment. Rather than impose broad restrictions on open-source models, they should target demonstrated risks and allow private investment and industry coordination to expand competitive American alternatives. America’s historic advantage has come from pairing invention with diffusion. It will capture AI’s full economic potential only if it does the same again.
Dr. Paul Prentice is a retired professor of Economics and Business and is a senior fellow with the Independence Institute
Read more ‘Uncle Bernie’ and the woman who would be president
Image: Easy-Peasy AI