Securing America's compute advantage: Anthropic’s position on the diffusion rule (www.anthropic.com)

Anthropic has formally responded to the U.S. Department of Commerce’s interim "Framework for Artificial Intelligence Diffusion" rule, advocating for strengthened export controls on advanced AI semiconductors to preserve America’s strategic compute advantage. Central to their position is the assertion that maintaining U.S. leadership in cutting-edge AI hardware is critical for national security and economic prosperity, especially as global AI competition, notably with China, intensifies. The Diffusion Rule, introduced in January 2025, implements a tiered system regulating chip exports to ensure allies have greater access while restricting adversarial nations, addressing the risk of compute infrastructure offshoring. Anthropic’s analysis underscores that advanced AI training depends heavily on leading-edge chips, which remain predominantly U.S.-developed. Chinese firms like DeepSeek have made progress but face significant bottlenecks due to export restrictions, resulting in notably higher operational costs and less efficient AI training. The submission recommends refining the framework by adjusting country tiers based on data center security, lowering the threshold of “no-license” chip purchases to close smuggling loopholes, and boosting enforcement funding. Anthropic warns that delays risk stockpiling and smuggling, undermining the control regime's effectiveness. Their proposals aim to secure America’s technological infrastructure and ensure that pivotal AI advancements and compute resources remain domestically anchored, safeguarding U.S. AI leadership in the crucial years ahead.