Shocking moves by the US government to restrict access to the best artificial intelligence systems from Anthropic and OpenAI have sparked growing interest in open source models – especially those from China.
A White House’s de facto bans against regulations blindsided the tech world, which had grown accustomed to AI labs churning out increasingly powerful models with little concern for government interference.
The episode highlighted a long-standing debate: open versus closed artificial intelligence.
Most of the most popular AI models—like OpenAI’s ChatGPT or Anthropic’s Claude—are “closed,” meaning the company keeps the underlying code and data locked away.
Users can access AI through an app or website, mostly through a subscription, but the company controls who gets in and can shut down access at any time.
“Open source” or “open weight” models work differently: developers release the core model files for anyone to download, modify, and run on their own computers. Once released, no one—not the company, not a government—can take them back.
In early June, the Trump administration ordered Anthropic to block non-Americans from using its most powerful and closed models, the Mythos 5 and Fable 5.
Faced with the complexity of user review, the startup simply pulled the models completely offline.
Soon after, OpenAI agreed to let the government approve any client for its newest model, GPT-5.6.
“If everything you have to do has to be in a specific boundary model, it makes what you’re building less reliable” when it’s suddenly unavailable, said Oren Michels, co-founder and CEO of Barndoor AI.
Haitham Mengad, co-founder of Stems Labs, a startup focused on creating AI-powered music, felt the disruption firsthand.
“Fable has been a game-changing model for me. Honestly, when they took it away, that was the first time I realized … it’s almost like a drug,” he recalls.
The Mythos episode “was a powerful moment” to look at open source as an alternative, Mengad said.
– ‘Being flexible’ –
Open models were already gaining fans because using closed AI continues to become more expensive.
Around the same time, China’s Zhipu AI (also known as Z.ai) released GLM-5.2, an open model that performed almost identically to the flagship offerings from Anthropic and OpenAI on several benchmarks.
“GLM-5.2 is free to download, patch and run on an enterprise’s own servers, putting pressure on prices in frontier labs at the same time the approach looks shaky,” noted AI analyst Andrew Curran.
On OpenRouter, a platform that routes requests across different AI models, the combined usage share of Google, Anthropic and OpenAI fell from 55 percent to 33 percent between January and June.
China’s Open DeepSeek now leads by a clear margin.
“You want to be as flexible as you can be. Maybe a year and a half ago some big company could say we bought Anthropic or we bought OpenAI, and now nobody, nobody buys just one,” Michels said.
Among Western companies, France’s Mistral stands largely alone in championing open designs. American tech giant Meta, once a vocal advocate of open source, has backed away from this.
Meanwhile, early doubts about Chinese AI designs as a security threat are fading, at least somewhat.
“I don’t think there’s any danger, to be honest,” Mengad said. Fear is more “psychological, emotional than rational”.
Once you download an open source model and use it on your hardware, the company that made it—Chinese or otherwise—has no access to your data or control over how you use it.
However, some experts think the government crackdown may also end for open models as they become more powerful.
“If Mythos-level models are considered dangerous, China won’t want them open either,” said Ethan Mollick, a professor at the University of Pennsylvania and a leading voice on AI — meaning that governments everywhere, not just Washington, may want to keep high-level AI under wraps.
