The Trump administration’s move to limit deployment of top U.S. AI models has opened an unexpected lane for Chinese developers to gain ground. After a two-week halt, Anthropic received partial clearance to release its Mythos 5 to select users while Fable 5 remains on ice; OpenAI is also slowing GPT 5.6 at Washington’s request. Meanwhile, China’s Zhipu is touting GLM 5.2, which researchers and investors say matches some leading U.S. capabilities on cybersecurity benchmarks at a fraction of the cost, encouraging U.S. firms to trial cheaper, open-weight alternatives such as DeepSeek and Kimi K2.7. With corporate buyers pivoting from “tokenmaxxing” to return on investment, lower-cost Chinese models are gaining traction at companies like Shopify, Airbnb and Coinbase. The shift raises fresh national-security and compliance concerns as open-weight models make it easier to run powerful systems outside major cloud providers and potentially automate more of the cyberattack chain.
