Frontier AI Access Controls Expand to Cloud Compute
Key Questions
Which Chinese open-weight models are gaining traction?
Kimi K3 from Moonshot AI, Zhipu GLM-5.2, DeepSeek, Z.ai, and Inkling are rivaling US models, with Moonshot halting new Kimi K3 subscriptions due to high compute demand.
What security concerns do Chinese open-weight models raise?
These models increase risks around model distillation, supply chain security, and geopolitical attack surfaces, especially as they are adopted by US companies like Airbnb and Pinterest.
How has Alibaba contributed to open-source AI?
Alibaba open-sourced its SAIL platform to challenge Nvidia CUDA dominance and further expand the influence of Chinese AI technologies.
What action did South Korea take regarding DeepSeek?
South Korea banned DeepSeek over data privacy concerns, highlighting international tensions around Chinese open-weight AI models.
What position have 25 US tech companies taken on open-weight models?
Nvidia, Meta, Microsoft and others jointly lobbied Washington to embrace open-weight models and defend model distillation, directly challenging the Gold Eagle framework.
The Gold Eagle clearinghouse is restricting distribution of frontier models, including Claude Mythos 5 and Fable 5. Proposed controls on Chinese firms' overseas cloud capacity and Nvidia chips, Anthropic's reported $35 billion compute commitment, and Europe's $451 million AMD LUMI-AI deal expose difficult KYC, jurisdiction, attribution, supply-chain, and concentration problems.