Anthropic said the US can cement a one- to two-year AI lead against China — but only if it acts fast.
The company laid out two possible scenarios for the AI landscape in 2028 in a lengthy Thursday post: one in which the US restricts China’s access to American AI compute, and another in which it does not.
Anthropic said China is closing the AI gap with the US through loose controls on chip exports and through distillation attacks, which involve using a developed AI model to train a smaller “student” model.
The company wrote that “if the US and its allies act now to address both issues, it may be possible to lock in a 12-24 month lead in frontier capabilities.”
Anthropic added that “the window of opportunity to lock in that lead will not necessarily remain open for long.”
The US gaining a lead will be crucial to making AI safe, the company said, adding that a “neck-and-neck race between American and Chinese AI labs could make industry and government-led safety and governance efforts more difficult.”
With such tight competition, AI labs in both countries feel more pressure to release new models faster without properly testing their safety measures, it added.
Anthropic called for policy changes to strengthen chip export controls, ramp up enforcement budgets, and stop distillation attacks by Chinese AI labs.
“Our past success means that our present task is largely to avoid squandering our advantage: to decide not to make it easier for the CCP to catch up,” the company wrote, referring to the Chinese Communist Party.
In February, Anthropic said in a statement that three of China’s biggest AI companies, DeepSeek, MiniMax, and Moonshot AI, were “illicitly” using Claude to further their own models.
The Biden administration first imposed export controls on American chips to China in 2022. The Trump administration has built upon those regulations by prohibiting Nvidia and AMD from selling chips to China. But last August, it partially backtracked on this decision, allowing Nvidia to sell its H200 chips if it pays the US government a 25% levy on its sales.
And underground operations by Chinese actors seeking access to American chips persist. In December, the US charged several parties for attempting to smuggle Nvidia’s most advanced chips by relabeling them as “SANDKYAN” chips.
Contrary to Anthropic’s assertion that China is closing the AI gap, ex-ByteDance engineer Zhang Chi said in April that it’s actually falling further behind.
Zhang, who is now a research scientist and assistant professor at Peking University, said in a podcast interview that Chinese AI lacks high-quality data to train its models, as well as access to advanced chips.
Anthropic’s post came on the same day that President Donald Trump met with Chinese leader Xi Jinping, the first time a US president has visited China since Trump’s 2017 trip. Trump brought an entourage of US business leaders, including Tesla CEO Elon Musk, Apple CEO Tim Cook, and Nvidia CEO Jensen Huang.
