Key facts
- China is considering restricting foreign access to its advanced AI models.
- Discussions involved treating cutting-edge AI as a critical national asset.
- Potential measures include classifying AI theft as a national security offense.
- Chinese startup DeepSeek is developing its own AI chip for inference.
- This move aims to reduce DeepSeek's reliance on Nvidia and Huawei chips.
- U.S. export controls on advanced chips to China are a factor in DeepSeek's strategy.
Chinese authorities are reportedly considering measures to restrict foreign access to the country's advanced artificial intelligence models, treating them as a critical national asset. These discussions, involving major tech firms like Alibaba and ByteDance, mirror national security concerns and controls previously implemented by the U.S.
Potential measures include classifying AI technology theft as an offense under national security law and restricting funding for domestic AI startups. The scope of these restrictions, which may apply to future models, is still under discussion.
In parallel, Chinese AI startup DeepSeek is developing its own inference chip, aiming to reduce its dependence on foreign hardware from companies like Nvidia and Huawei. This move follows global trends among AI developers seeking greater control over their hardware and aligns with China's push for domestic technological alternatives amid U.S. export controls.
DeepSeek, known for its globally popular AI models, has relied on Nvidia's H800 and Huawei's Ascend chips. The company is now engaging external partners and increasing hiring for chip design engineers, though the effort is in its early stages. Success in chip design and manufacturing faces significant hurdles, including U.S. restrictions on access to advanced foundries and components.