China’s open-weight AI model Kimi K3 is an impressive piece of engineering, the biggest of its kind ever released, running rings around open-weight rivals on coding and priced to undercut Silicon Valley.The White House has accused its maker, Moonshot AI, of distilling it from Anthropic’s Fable, although officials have offered no evidence and sceptics have reasonable doubts. The pattern is better evidenced. Anthropic said in February that it had traced more than 16 million unauthorised Claude exchanges to Chinese labs. In June, it told the US Senate that Alibaba’s Qwen lab ran a further 28.8 million exchanges in six weeks.Subscribe to gift this articleGift 5 articles to anyone you choose each month when you subscribe.Subscribe nowAlready a subscriber? Fetching latest articles
Relying on China’s subsidised AI models is not smart policy
On critical technologies, Beijing subsidises market share, assists industrial-scale IP theft, and shuts its markets to competitors. Its AI strategy is similar.
TL;DRAI
Moonshot AI released Kimi K3, the largest open-weight model, underpriced vs Silicon Valley; Anthropic traced 16M+ unauthorized Claude accesses from Chinese labs. Cheap Chinese models threaten IP security and Western open-source market share in enterprise AI budgets.
121 words~1 min read
