The next phase of the US-China artificial intelligence race may be decided within two years, according to a new policy paper from Anthropic. The AI company warns that decisions made in 2026 could determine which nation controls the next generation of AI by 2028.
Anthropic’s report, titled “2028: Two scenarios for global AI leadership,” comes as President Donald Trump visits China alongside top Silicon Valley executives including Elon Musk, Tim Cook, Jensen Huang, and Larry Fink. The high-profile delegation underscores the geopolitical stakes surrounding AI development.
Chips and Compute at the Center of the Conflict
Anthropic argues that access to advanced chips and semiconductor manufacturing tools is the primary bottleneck for China’s AI progress, not talent or research capability. Chinese firms such as Huawei, Alibaba, and ByteDance have remained competitive by exploiting export-control loopholes, accessing overseas data centers, and using distillation techniques to replicate American AI models.
The paper specifically mentions DeepSeek, a Chinese AI lab that trained its models on advanced US chips technically restricted from sale to China. This highlights the ongoing challenge for US export controls, which are designed to limit China’s access to cutting-edge hardware but are frequently bypassed through smuggling, third-party transfers, and remote computing access.
Anthropic describes “distillation attacks” as another major concern. Chinese labs create fraudulent accounts to extract outputs from American models like those from OpenAI, Google, and Anthropic itself, then replicate their capabilities at reduced cost. The Frontier Model Forum, an industry group, has publicly criticized this practice.
Two Scenarios for AI Leadership
The report presents two starkly different futures. In the first scenario, the US and allied democracies maintain their lead by tightening export controls, limiting distillation attacks, and accelerating global adoption of American AI technology. “In this world, democracies set the rules and norms around AI,” Anthropic wrote.
In the second scenario, if Washington fails to close loopholes around chip exports and remote compute access, Chinese firms could catch up or surpass US companies. “In this world, AI norms and rules are shaped by authoritarian regimes, and the best models enable automated repression at scale,” Anthropic warned.
AI as a Military and Political Tool
Anthropic frames AI not just as an economic driver but as a national security technology. Frontier AI systems could reshape cyber warfare, military planning, and surveillance capabilities. The Chinese government already uses AI for censorship, monitoring, and cyber operations; advanced systems could dramatically expand these abilities.
Anthropic also cautions that a close AI race could weaken safety standards, as companies and governments feel pressured to release increasingly capable systems faster. The report notes that only a few Chinese AI labs publicly disclose safety testing results for high-risk areas like chemical, biological, radiological, and nuclear threats. This opacity raises concerns about global stability if China achieves a leadership position.
Policy Recommendations and Historical Context
The paper arrives amid intense debate in Washington over semiconductor export restrictions. U.S. policymakers are weighing how much access Chinese firms should have to advanced Nvidia hardware, which is critical for training large AI models. Anthropic urges the US to tighten controls on advanced chips, crack down on smuggling and overseas compute access, and expand efforts to promote American AI systems worldwide.
Historically, the US has enjoyed a significant lead in AI research and development. The country is home to many of the world’s top AI labs, including OpenAI, Google DeepMind, and Anthropic itself. However, China has invested heavily in AI as part of its national strategy, with the goal of becoming the world leader by 2030. The narrowing gap has made the race increasingly competitive.
Anthropic claims that “there is a high likelihood that we will look back on 2026 as the breakaway opportunity for American AI.” The company supports continued dialogue with Chinese experts on AI safety but emphasizes that its concerns are directed at the Chinese Communist Party, not the Chinese people or the broader research community.
The potential consequences of losing the AI race are severe. If China achieves leadership, the global adoption of AI could be shaped by authoritarian norms, leading to enhanced surveillance, repression, and limitations on free expression. Conversely, US leadership could promote democratic values, transparency, and safety standards.
Anthropic’s paper ultimately frames 2026 as a narrowing window for US policymakers. Whether its predictions prove prescient or self-interested, the company argues that decisions made now will shape not only who builds the strongest AI systems but also whose rules govern their use. The race is not just about technology; it’s about the future of global governance.
Source: eWEEK News