2028: Two scenarios for global AI leadership

Anthropic Research released a paper that presents two possible scenarios for the state of U.S.-China AI competition in 2028. The authors argue that the United States and its allies must maintain a compute advantage by tightening export controls on advanced chips and disrupting China’s large‑scale distillation attacks that replicate U.S. innovations. They contend that if the U.S. fails to act, Chinese labs could catch up to or surpass U.S. frontier models, allowing the Chinese Communist Party to use AI for large‑scale repression and to shape global AI norms. In the first scenario, democracies retain a commanding lead, set rules and norms, and can engage China on safety. In the second scenario, authoritarian regimes dictate AI development, enabling automated repression. The paper emphasizes that the political system that builds the most capable AI will determine the rules, safety, and interests served by the technology, and that democratic governance is essential for a safe transition to transformative AI. The authors note that the window to lock in a 12‑24 month lead is urgent and that the U.S. and allies have historically succeeded by enforcing export controls and protecting intellectual property.