America Trembles as China's AI Push Intensifies

Chinese technology once sent shockwaves through Silicon Valley with the introduction of DeepSeek’s R1 AI model, which was developed at a low cost yet rivalled the performance of advanced US-made models. At the time, shares of American technology giants tumbled in unison, triggered by investor concerns over expensive AI models that had yet to deliver tangible profits. Since the emergence of DeepSeek, Chinese technology has continued to develop rapidly, leaving the US on edge. Most recently, China made another breakthrough with the launch of the ‘Kimi K3’ model, which recorded stellar performance on several important benchmarks. In some benchmarks, Kimi K3 was able to match leading frontier models from the US, which were developed at exorbitant costs. The launch has sparked a fresh wave of panic that China is closing the gap with the US in the AI race. The US continues to accuse advanced Chinese AI models of being trained on the backbone of American frontier models from firms like Anthropic, OpenAI, and Google. Amidst these tensions, a crucial strategic difference has emerged between AI development in China and the US. China is implementing open-source and open-weight models, whilst the US continues to maintain closed-source models. The difference lies in the accessibility of the source code. Open-source models allow developers to freely view, modify, and redistribute the available AI model, whereas closed-source models keep the code secret, protected by copyright, and only modifiable by its creator. Open-weight models sit somewhere in the middle, where the final parameters of a trained model are published for anyone to download and use locally. Simply put, open-source hands over the entire ‘recipe’ for the AI model, while open-weight only provides the finished product. ‘I am personally surprised that China continues to allow such high-quality models to be open-sourced, given the potential risks,’ wrote Dean Ball, a former senior AI advisor to President Donald Trump who now serves as head of strategy at OpenAI, on platform X. He argued that an open-weight strategy would lead to full ‘AI communism’, whilst open-source could slow progress by reducing capital expenditure on AI. However, what really ignited the discussion was his subsequent statement: ‘I predict that at some point the Trump administration will realise that their best strategy here is to create significant regulatory risk around the use of Chinese open-weight models.’ Ball suggested that creating fear, uncertainty, and doubt through the regulatory process would cause most US companies to avoid using such open-source models. He later clarified that this was a prediction, not a recommendation, and affirmed his support for open-source models as long as AI has not become too dangerous. Anthropic and OpenAI insist that their models are too advanced for public access, arguing that doing so is dangerous as it allows anyone to use their tools for any purpose with minimal oversight. Closed-source models give creators greater control, including over security, access, and pricing. These labs have warned that Chinese open-weight models pose a threat to both national security and their businesses. David Sacks, a venture capitalist who served as the first White House AI and crypto czar under the Trump administration, said that ‘weaponising regulatory uncertainty is completely unacceptable.’ Responding to Ball’s post, he wrote: ‘We are at a critical inflection point in AI policy. The leading labs with closed-source systems, which have formed a duopoly in model revenue, want the government to eliminate their open-source competitors.’ He identified the duopoly as OpenAI and Anthropic, adding: ‘They’ve shown their intentions openly. Now it’s time for the rest of Silicon Valley, the vast majority who still believe in open competition, to do the same.’