White House AI crackdown opens door for Chinese model makers to close gap

• Zhipu's GLM 5.2 reportedly matches Anthropic's capabilities on some cyber benchmarks and costs one-quarter the price per token, prompting companies like Coinbase to adopt it and cut AI spending by nearly half
• Anthropic's Fable 5 model remains off the market after a two-week shutdown due to export control directives, while OpenAI also limited its rollout following government requests
• Open-weight Chinese models are easily accessible to U.S. users who can download and run them on their own servers, with cybersecurity experts warning these models could automate entire cyberattacks within months

Summary: White House AI Restrictions May Benefit Chinese Competitors

The Trump administration's crackdown on advanced U.S. AI models from Anthropic and OpenAI could allow Chinese companies to narrow the technological gap. After a two-week shutdown due to export controls, Anthropic received limited permission to resume operations Friday, though its Fable 5 model remains unavailable. OpenAI also restricted its rollout following government requests.

Key Development: Chinese company Zhipu launched its open-weight model GLM 5.2 this month, which researchers claim matches leading U.S. labs on some cybersecurity benchmarks, potentially equaling capabilities of restricted American models. Venture capitalist Marc Andreessen called it "the first Chinese AI model to match and often beat the American big lab public AI models."

Market Implications: Jefferies strategist Christopher Wood notes GLM 5.2 "is almost equal to Anthropic as a competitor for the corporate market" at just one-quarter the cost per token. This arrives as corporate America shifts from unlimited AI spending ("tokenmaxxing") toward efficiency and ROI focus.

Adoption Trends: Major companies are already embracing Chinese alternatives:

• AI startup Lindy moved 100% of traffic from Claude to DeepSeek, dramatically reducing costs
• Coinbase CEO Brian Armstrong reported cutting AI spending nearly in half using models like GLM 5.2 and Kimi 2.7
• Companies including Alibaba and ByteDance have deployed China's Qwen 3

Security Concerns: Experts warn open-weight models enable easier cybersecurity threats, with some capable of automating cyberattack stages. Former Trump AI czar David Sacks suggested the restrictions contradict the president's pro-innovation AI strategy.

The situation represents a potential strategic setback as U.S. export controls inadvertently create opportunities for Chinese AI advancement.