Anthropic warns China could surpass the US in AI race by 2028 without chip controls

Anthropic released a policy paper warning that China could surpass the US in AI development by 2028 if Washington weakens chip export controls, citing loopholes, overseas compute access, and state-backed investment. The company argues advanced AI will become a critical geopolitical tool, shaping military, cyber, and economic power, and urges stricter semiconductor restrictions to maintain a US lead.
Anthropic, an AI research company, has published a policy paper asserting China could overtake the US in frontier AI development by 2028 without stronger chip export controls. The report highlights that current restrictions have slowed China’s progress but warns of persistent loopholes, including overseas compute access, smuggling operations, and 'distillation attacks' that replicate US AI models. The paper frames advanced AI as a strategic technology influencing military power, cyber warfare, and global influence. Anthropic argues that authoritarian governments may exploit AI for mass surveillance, censorship, and offensive cyber operations, while democratic nations risk falling behind if they fail to enforce semiconductor restrictions. Advanced semiconductors remain central to AI leadership, with US firms like NVIDIA dominating the market for AI GPUs. Anthropic claims China’s state-backed investment and engineering talent are already competitive, but compute access remains the primary barrier to matching US AI labs. The company estimates that stricter enforcement of chip controls could extend the US lead by up to two years by the end of the decade. Two scenarios for 2028 are outlined: one where the US maintains dominance, shaping global AI standards, and another where China closes the gap, potentially altering technological and geopolitical landscapes. Anthropic emphasizes that decisions made before 2028 will determine whether democratic or authoritarian governments dictate the future of AI development.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.