The Trump administration’s June order forced Anthropic to block non‑American users from its most powerful closed models, Mythos 5 and Fable 5. Rather than build a complex screening system, the startup pulled the models offline entirely. Shortly after, OpenAI agreed to submit every customer for its newest GPT‑5.6 model to government approval, highlighting a new era of regulatory uncertainty for frontier AI.

Closed vs. Open AI

Well‑known models such as OpenAI’s ChatGPT and Anthropic’s Claude are “closed”: the underlying code and training data remain proprietary, users access the service via apps or subscriptions, and the provider can revoke access at will. Open‑source or “open‑weight” models, by contrast, release their core files publicly, allowing anyone to download, modify, and run them on personal hardware – a move that cannot be reversed by a company or a government.

Impact of the U.S. Crackdown

Barndoor AI co‑founder Oren Michels warned, “If everything you need to do depends on a specific frontier model, your project becomes far less reliable when that model disappears.” The sentiment is echoed by Haitham Mengad of Stems Labs, who described the removal of Fable as “almost like a drug” for his AI‑driven music‑creation workflow.

China’s Open‑Source Surge

At the same time, China’s Zhipu AI (Z.ai) released GLM‑5.2, an open model that matches Anthropic and OpenAI on several benchmarks. Because GLM‑5.2 can be downloaded, fine‑tuned, and run on an enterprise’s own servers for free, it puts pricing pressure on Western labs whose access now looks shaky. AI analyst Andrew Curran noted, “Open‑source availability is forcing frontier labs to compete on cost as well as capability.”

Shifting Market Share

Data from OpenRouter shows the combined usage share of Google, Anthropic and OpenAI fell from 55 % in January to 33 % by June, while China’s DeepSeek has taken a clear lead. “A year and a half ago a large company might say we bought Anthropic or OpenAI; today no one buys just one,” said Michels, underscoring the growing desire for flexibility.

Future Risks and Regulation

While many Western firms, such as France’s Mistral, still champion open models, others like Meta have retreated from open‑source advocacy. Some experts warn that as open models become as powerful as Mythos‑level systems, governments worldwide may seek to restrict them. University of Pennsylvania professor Ethan Mollick cautioned, “If Mythos‑level models are deemed risky, China will also be reluctant to keep them open.”

In summary, U.S. restrictions have reshaped the AI landscape, accelerating the shift toward open‑source alternatives and positioning China as a pivotal player in the next wave of generative AI.