Banning Open-Weight AI Models Could Cost America the AI Race | RealClearMarkets
Subscribe
Sign In | Subscribe
up next:
now reading:<br>Banning Open-Weight AI Models Could Cost America the AI Race
Banning Open-Weight AI Models Could Cost America the AI Race
By
Jon Decker
August 13, 2026
Markus Spiske
Markus Spiske
Story Stream
recent articles
Video: Global Warming Lorem Ipsum Dolor Sit ...
Article: Global Warming Lorem Ipsum Dolor Sit ...
Article: Global Warming Lorem Ipsum Dolor Sit ...
Entry: Global Warming Lorem Ipsum Dolor Sit ...
Video: Global Warming Lorem Ipsum Dolor Sit ...
America will win the race to superintelligence unless we defeat ourselves. Unfortunately, there is a scenario where that happens — and as a policy matter, it’s already under consideration. The bans and restrictions on open-weight AI models floated by politicians and Anthropic alike are exactly that scenario.
To give a brief primer, an open-source AI model is one whose code is made publicly available to anyone, anywhere in the world, so that they may adapt it as they wish. Motivations for doing so include fixing errors within the AI system or retooling it so that it can be specialized to carry out specific tasks for an individual or corporation.
On the other hand, a closed-source model is one whose code is kept proprietary to protect the technology that powers it. Many of the leading AI companies in America take this approach with their frontier models under the belief that it is more advantageous to their profit margins.
The middle-ground is an open-weights model, in which underlying training data and proprietary recipes are kept secret, but the model architecture code and parameter weights are made publicly available for anyone to run, modify, and build on.
There is nothing inherently wrong with any of these models; they all have their advantages and disadvantages. However, one of the trends we are seeing right now in the AI space is that China’s release of prominent, mainly open-weight models has allowed them to close the gap on America’s leading frontier models at a speed that has taken our largest players by surprise.
The reason for this can best be illustrated by paraphrasing my Parkview Institute colleague, John Tamny. Tamny has previously used the analogy that "while you may be the smartest person in a football stadium, collectively, the stadium is a lot smarter than you." Applied to the debate over open v. closed models, China is effectively crowdsourcing improvements on its AI systems to the entire globe, whereas many American companies are strictly relying on their in-house talent to compete. As smart as the employees of major AI players undoubtedly are, one can see how competing against the global crowd is proving increasingly challenging.
Where this story takes a startling turn is that some have drawn completely the wrong conclusion from this turn of events. Most prominently, Anthropic has seemingly adopted an "if you can't beat them, ban them" mindset, where its solution is to restrict the usage of open models as a whole. Where this approach falls flat is that nothing about banning or restricting open models domestically will cause Beijing to stop developing them.
Even if such usage were restricted in America, China would still have the rest of the world improving its technology — except America’s independent researchers and computer engineers would be deprived of the ability to analyze and learn from those models.
Considering the serious attention Anthropic’s misguided approach to public policy has received, it came as much-welcomed relief that leading companies in the AI space —including Nvidia, Microsoft, Meta, Alphabet, and OpenAI — recently signed a letter urging policymakers to avoid heavy-handed regulations on open-weights models. Elon Musk also voiced his "full support" on X.
In addition to sounding the alarm on how restricting open models would reduce American competitiveness, the letter made the critical point that closed models aren’t inherently safer (as those calling for a ban would like you to believe). Recently, when OpenAI’s frontier model was caught breaking out of its sandbox to launch a cyberattack on another company, it was an open-weight model that helped contain the intrusion.
Some believe open-source models are more dangerous because they allow users to change the code, but those changes often include fixes that strengthen the security of the model. If an individual company suffers from "tunnel vision" that overlooks certain vulnerabilities, the rest of the world can help fix them.
For America to dominate the AI field, it is critical that we strive for leadership in both open- and closed AI models. Banning competition doesn’t make it go away.
To quote the preeminent supply-side economist Robert Mundell: "The only closed economy is the world economy." This is particularly true for AI.
Jon Decker is the Executive...