Anthropic: US Can Lock In 12–24 Month AI Lead Over China by 2028
Anthropic published a research paper arguing the US and allies can lock in a 12–24 month frontier AI lead over China by 2028 by restricting Chinese access to advanced compute and blocking distillation attacks. The paper frames compute as "the gatekeeper for training, deployment, revenue, and future model improvement." China is alleged to use loopholes, smuggled chips, and offshore data centers to stay close to US frontier labs. Critics including Yann LeCun and Dan Jeffries characterized the paper as a "21st century East India Company whitepaper."
Why It Matters
The paper is Anthropic's most direct intervention in AI geopolitics. The compute-gating argument drew concerted pushback from the open-source and academic AI community, creating the week's dominant discourse thread across AI-focused accounts.