Anthropic published a new policy paper that pulls no punches: if the US doesn’t tighten semiconductor export controls, China could close the AI gap by 2028 — faster than many policymakers expect.
Two scenarios for 2028
The paper describes two possible futures. In the optimistic scenario, the US and its allies maintain their lead in frontier AI. Democratic nations shape global AI standards, safety practices, and infrastructure deployment.
In the pessimistic scenario, weaker enforcement of chip controls enables Chinese firms to close the gap with American AI companies. Beijing would expand its influence over global AI infrastructure.
The core argument
Anthropic’s point is straightforward: advanced chips are the bottleneck. Whoever controls the chips controls AI development. And since the most important chips are still designed by American and allied companies, the US currently has a strategic lever — but the lead is eroding.
The timing is politically charged: the paper dropped on the same day that US Treasury Secretary Bessent met with Chinese representatives to discuss AI rules. And OpenAI proposed a global AI oversight body modeled on the International Atomic Energy Agency on the same day.
Context
Anthropic has been taking a clear stance on China for months. Back in February, CEO Dario Amodei accused Chinese AI labs of systematically extracting Claude’s outputs. The new paper provides the strategic framework for that position.
For the AI industry, the geopolitical dimension is becoming increasingly important. It’s no longer just about building better models — it’s about who writes the rules. And Anthropic wants to make sure it’s the democracies.
Sources: Anthropic Research, Interesting Engineering, China Money Network