U.S. AI Export Limits Hand Advantage to Chinese Competitors

Export controls collide with Chinese AI advances

The Trump administration's recent export restrictions on leading U.S. artificial intelligence models have created an opening for Chinese competitors at a critical moment in the global AI race. After a two-week shutdown, Anthropic received White House approval Friday to release its Mythos 5 model to select companies and federal agencies, though its Fable 5 model remains blocked. OpenAI announced the same day it would limit rollout of GPT 5.6 models following a government request.

The timing proved particularly awkward. Just as U.S. labs faced restrictions based on national security concerns, Chinese company Zhipu released its GLM 5.2 model earlier this month. According to researchers, the model performs on par with top U.S. labs on certain cyber benchmarks, matching some of Mythos' capabilities.

Why it matters

The development challenges a core assumption of U.S. AI policy: that limiting domestic regulation would help American companies outpace China. Instead, export controls intended to protect national security may be handing market share to the adversary they aim to contain—especially as corporate America shifts from unlimited AI spending to cost efficiency.

Market dynamics favor Chinese alternatives

Corporate adoption patterns underscore the competitive threat. Flo Crivello, CEO of AI startup Lindy, recently moved 100% of his company's traffic from Anthropic's Claude models to DeepSeek, a Chinese provider offering cheaper open-weight alternatives. "We did it, and you could see that cost curve go down, like, crash to the ground," Crivello told CNBC.

Jefferies strategist Christopher Wood wrote in a client report that GLM 5.2 "is almost equal to Anthropic as a competitor for the corporate market and is just one quarter of the cost in terms of cost per token," citing industry sources. Major companies including Shopify and Airbnb have publicly discussed using Alibaba's Qwen 3 model. Coinbase CEO Brian Armstrong disclosed his company uses open-weight models like GLM 5.2 and Kimi 2.7, cutting AI spending nearly in half despite increased usage.

Venture capitalist Marc Andreessen wrote on X that "many smart people/AI insiders are saying GLM-5.2 is the first Chinese AI model to match and often beat the American big lab public AI models with no compromises."

Open-weight distribution complicates enforcement

Chinese AI developers reach U.S. users easily because companies can download open-weight models and run them on their own servers without third-party cloud providers. "With the open-weight models, it's kind of the Wild West," said Travis Lanham, co-founder of AI security startup Armadin, which is testing GLM 5.2 and Moonshoot AI's Kimi K2.7.

Lanham reported the models show improved capabilities for cybersecurity applications including analyzing reconnaissance data and creating exploit code. Hed Kovetz, CEO of cybersecurity startup Silverfort, warned that some open-weight models can already automate many cyberattack stages and may be months away from running entire operations autonomously.

Former Trump crypto and AI czar David Sacks, typically a critic of Anthropic's safety approach, posted on X: "A year ago, President Trump declared that America was in a global AI race and that the way to win it was to be pro-innovation, pro-infrastructure, pro-energy, and pro-export. President Trump was exactly right; we deviate from that strategy at our peril."

Sam Bresnick, a research fellow at Georgetown's Center for Security and Emerging Technology, called the developments "a pretty good wake-up call."

These details were first reported by CNBC.