This week’s reporting out of Washington (See: “Top American AI Execs Sound Alarm on Chinese Models”, The Wall Street Journal, July 20) describes an administration weighing whether to restrict American access to open weight AI models, particularly the ones released by labs in China. 

The measures reportedly under discussion include trade block-lists, security warnings, and a possible executive order aimed at open models themselves. Given the emerging capabilities of these technologies, it’s sensible for us to seek a more mindful approach in how we develop and release them. But before we reach for restrictions, we should be clear about where this technology came from, how it actually works, and what walling it off would cost us.

A moment built in the open

Every AI system in the headlines today, whether proprietary or Open Source, exists because researchers shared their work openly. The transformer architecture at the heart of modern AI was published for anyone to read and build on. That openly shared research produced Open Source software: the frameworks used to train these models, the libraries used to evaluate them, the operating systems, orchestration layers, and inference engines used to run them. When a frontier lab trains a closed, proprietary model, it does so on a stack of Open Source software. The same is true of every open model. 

There is one foundation under this whole industry, and it was built by people and organizations collaborating in the open.

The scale of that foundation is easy to underestimate. In our response to the White House AI Action Plan, we cited estimates that companies would need to spend roughly 3.5 times more than they do today, nearly $9 trillion, to rebuild what Open Source software already gives them. Harvard-backed research puts the demand-side value of Open Source at $8.8 trillion. 

Open Source is economic infrastructure. Policy that treats it as a market niche, or as somebody else’s product, will get the big decisions wrong.

A wall keeps out more than it keeps in

Open collaboration has never checked passports. The methods, weights, and tooling of Open Source AI circulate through a global community, and improvements flow in every direction: a technique published in Beijing gets refined in Boston, stress-tested in Berlin, and deployed in Bangalore within weeks. Cutting American developers off from openly released models would not freeze that exchange. It would simply continue without them. Researchers everywhere else would keep studying, distilling, and improving on the strongest open models, while practitioners in the U.S. would be legally fenced out of work the rest of the world treats as a shared starting point.

Worth noting is the surrender of something that security-minded policymakers should care about: visibility. Openly available weights can be examined, tested, and benchmarked by anyone, including American security researchers and government evaluators. Restrictions do not make those models disappear. They make them opaque to us while leaving them available to everyone else. 

Startups will pay the bill

If restrictions like the ones under discussion take effect, the costs will not fall primarily on the frontier labs in the headlines. They will fall on the American startups and enterprises that have built on open models precisely because they can control them.  A coalition of more than 20 big tech companies and an association of 200 small companies are urging the Trump administration not to restrict access to Chinese open-weight AI models, warning that doing so could hamper the next generation of U.S. innovation. Running an open model means your data stays on your infrastructure, under your security controls, subject to your audits. You can pin a version, inspect its behavior, fine-tune it for your task, and run it disconnected from the internet entirely.

For many organizations, that control is the answer to the security challenges at the heart of this debate. And the research we cited in our Action Plan response found no evidence that open weight models are more vulnerable to misuse than closed systems.

Take those options away and the alternatives get narrower and more expensive. Startups that assemble products from affordable, modifiable components would be pushed toward a handful of proprietary APIs, with pricing set by the very incumbents advocating for restrictions. As we told the White House in our comments on the national AI R&D strategy, confining AI development to proprietary systems disadvantages smaller competitors and researchers while international rivals gain ground. Openness has been a pillar of American technological leadership for decades. It is a strange moment to abandon it.

Compete on value, not on exclusion

None of this requires anyone to lose. American frontier labs make products of extraordinary capability, and they have every legitimate way to compete: better models, better reliability, better integration, better service, better trust. Companies that believe their offerings are worth a premium should make that case to customers. What they should not do is ask the government to box the global Open Source ecosystem out of the U.S. market, and then call the result competition.

There is a better policy path, and much of it already exists. We worked with the G7 on its Vision on AI Openness, which gives governments shared language for degrees of openness. This vision states clearly the importance of AI openness:

AI openness has been an essential contributor to our economies, fostering innovation and cooperation, and broadening access to technologies for companies and communities.

Labeling, evaluating, and understanding open models is a far stronger security posture than pretending we can exclude them. The Open Source AI Definition gives everyone, including regulators, a shared framework for what openness in AI actually means: the freedom to use, study, modify, and share an AI system. The G7 vision shows governments and the Open Source community can write workable frameworks together. Invest in evaluation research, in open datasets, in the security of the commons, and in the people who maintain it. Those measures strengthen America’s position. Exclusion just isolates it.

Open Source works because people and organizations come together and collaborate to solve shared problems. That is how this technology came to exist, and it is how the United States has led before.