Anthropic publishes position piece on AI leadership by 2028

Anthropic published an essay, “2028: Two scenarios for global AI leadership,” arguing that the outcome of the US-China competition over frontier AI would likely be settled within a few years and that compute access, not raw model capability, was the deciding variable. The company laid out two futures: one in which the US and allied democracies preserve their lead by tightening export controls, disrupting Chinese firms’ extraction of capability from US models through so-called distillation attacks, and expanding global adoption of American AI systems; and one in which loopholes in chip export enforcement — including smuggling and remote access to restricted hardware — let Chinese labs reach parity or overtake the frontier.

Anthropic estimated that with tighter enforcement the US could hold roughly an elevenfold compute advantage over China’s AI sector, and it named DeepSeek as a case where restricted US chips had already been used to train an advanced model.

In this world, AI norms and rules are shaped by authoritarian regimes, and the best models enable automated repression at scale.

— Anthropic, “2028: Two scenarios for global AI leadership”

Anthropic framed the current period as decisive, writing that there was “a high likelihood that we will look back on 2026 as the breakaway opportunity for American AI” — a claim advanced by a company with a direct commercial stake in the policy conclusions it was arguing for, including continued and tightened export controls.