When venture capitalist David Sachs becomes the Trump administration's AI “czar,” he will face a thorny problem: Despite U.S. restrictions on chips and capital, China will continue to develop are Competent AI model — and share them openly.
The model is the inference engine that powers tools like ChatGPT. Once released openly, developers are free to modify it for different applications. Chinese models such as Alibaba's Qwen, 01.AI's Yi, and DeepSeek are currently most capable, most preferred and maximum Open your models globally and quickly establish global standards for AI.
These could be signs of a new Chinese AI strategy that uses open technologies to create global dependence on Chinese talent, capital, and industry.
That should be a cause for concern and a call to action. But U.S. policymakers are focused on the wrong threats. they are is reacting Alarmed by the prospect of China achieving AI equivalent to the Frontier AI Institute, prompt “Manhattan Project” accelerates the development and implementation of cutting-edge AI capabilities new controls It has the most powerful chip and the largest closed model, keeping these features out of Chinese hands.
But these efforts miss the point. The national security risk is not just that China could compete with frontier labs like OpenAI and Anthropic. The most immediate threat is that China is openly sharing capable AI models that could eventually underpin AI infrastructure around the world.
Open models are popular because they overcome major barriers to large-scale adoption of AI, such as trust, privacy, security, and cost. Developers can modify the open model to improve performance for specific tasks. Test, improve, and operate your AI systems without sharing sensitive data about your users, customers, patients, or citizens with your organization. widely distrusted A large tech company developing a closed-source model. Deploy AI systems independently and securely through your own on-premises hardware or compute provider. You can also build AI systems using a variety of competitive models without paying ongoing fees to a small number of frontier AI labs. These are some of the reasons why open models like Meta's Llama have surpassed 350 million downloads since 2023.
If the U.S. withdraws from open innovation in AI, other countries could “decouple” from U.S. technology and invest in their own infrastructure or look to alternatives to build satisfactory AI capabilities. be. That would entrench Chinese AI models such as Qwen, Yi, and DeepSeek, establishing global dependence on Chinese research labs and undermining the influence of American companies.
For example, Alibaba's model today is: downloaded Millions of times per month by developers. These models are embedded with different values (for example, the Chinese model is Required Laws and Alibaba's model that adhere to Communist Party doctrine exhibit signs of political censorship), and those values can shape the technology stacks of countless companies and governments around the world. Imagine a future where Baidu replaces Perplexity or Google and powers the AI search engine used by billions of people around the world.
Unfortunately, while some policymakers support open innovation in AI, others support restrictions on open technologies. The Federal Telecommunications Administration, refused Senators ran a bipartisan bid to support the restrictions. suggestion For licensing regimes that limit access to capable AI models. the house has progressed law With AI models subject to export controls, the current administration is preparing to impose further restrictions on the computing hardware needed to train, modify, and run capable models. And while the next vice president is here, I believed Although he has frequently criticized the “woke” values embedded in Big Tech companies' AI systems in support of open models, the president-elect has Expand controls It assessed key technologies as part of its trade policy toward China.
These measures risk stifling domestic AI innovation by cooling or hindering the release of open models, while at the same time accelerating the uptake of Chinese technology overseas. U.S. policymakers believe that AI leadership means more than just building advanced technology behind paywalls, but rather accelerating the adoption of AI at home and reducing global dependence on U.S. industry. You need to realize that it is about establishing.
Here's how the incoming Trump-Vance administration can act to maintain America's open source advantage.
First, the United States must invest in open development. Training advanced AI models is expensive, and open models cannot be taken for granted. the next president suggestion Because U.S. sovereign wealth funds can foster open innovation by investing in the infrastructure, talent, and resources that support open technologies that serve the national interest. Other countries are already doing so. Saudi Arabia, united arab emirates, Canada and France Each has pledged up to $100 billion and other incentives for AI, including significant investments in open model development.
Second, the United States needs to refocus its policies not just on AI safety, but on the proliferation and adoption of AI. The Trump-Vance administration should develop and implement a unified national strategy to promote open innovation in the field of AI. This may require a specialized organization to coordinate efforts across government and apply regulatory tools to foster private sector investment. For example, the U.S. Department of Energy and the Department of Commerce could streamline permitting and energy access for investments in open model developments. Additionally, federal procurement policy is likely to favor open models. sensitive Government applications such as defense and public services.
Finally, governments should avoid restricting the release of useful models, such as through export controls or licensing systems. Restricting the open model would limit global access to AI technology and leave a void for Chinese AI research institutes to fill. The open model brings unique challenges to AI monitoring. abuseand policymakers must remain vigilant against emerging risks. However, here is evidence of the devastating threat posed by open models. limiteddevelopers can apply layer of relaxation. Policymakers must balance these risks with the benefits of AI transparency and open models for competition, and the opportunity costs of ceding global AI leadership to China.
Rather than treating the open model as a threat, the new administration should see it as an opportunity. An open model can accelerate domestic AI adoption, shape global AI deployment, and establish lasting dependence on U.S. talent, capital, and industry. To win in AI, the United States must promote open source diplomacy.
Ben Brooks is a fellow at Harvard University's Berkman Klein Center and previously led public policy at Stability AI, the developer of Stable Diffusion.
Michelle Fang leads strategy projects at AI computing company Cerebras and previously organized Sam Altman's congressional hearings while on the Senate AI staff.
