The next phase of the US-China artificial intelligence rivalry could reach a decisive point as early as 2028, according to a new policy paper from Anthropic. The report, titled "2028: Two scenarios for global AI leadership," arrives as President Donald Trump visits China alongside Silicon Valley executives including Elon Musk, Tim Cook, Jensen Huang, and Larry Fink. Anthropic warns that policy decisions made today will determine whether democratic nations or authoritarian regimes control the future of AI.
Anthropic, one of the leading AI safety companies, presents two starkly different outcomes. In the first scenario, the United States and its allies maintain their lead by enforcing strict export controls, preventing Chinese firms from using distillation techniques to replicate American models, and accelerating global adoption of US AI technology. "In this world, democracies set the rules and norms around AI," the company wrote. In the second scenario, Washington fails to close loopholes around chip exports and remote access to computing infrastructure, allowing Chinese AI firms to 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.
Chips and Compute at the Center of the Fight
Anthropic's analysis places access to advanced semiconductors and computing power at the heart of the competition. The company argues that China's AI sector is constrained primarily by limited access to cutting-edge compute resources rather than by a lack of talent or research capability. Chinese firms have remained competitive by exploiting export-control loopholes, accessing overseas data centers, and using model distillation to mimic American AI systems. The report specifically names Huawei, Alibaba, and ByteDance as examples of companies advancing despite US restrictions. It also cites reports that Chinese models from DeepSeek were trained on advanced US chips that are technically forbidden from export to China.
Model distillation has become a major concern for US AI firms. Anthropic describes how Chinese labs create fraudulent accounts to extract outputs from American models and replicate their capabilities at lower cost. OpenAI, Google, and the Frontier Model Forum have all publicly criticized the practice. The paper argues that without stronger enforcement, these "distillation attacks" could erode America's technological advantage swiftly.
AI as a Military and Political Tool
Anthropic frames AI not only as an economic driver but also as a fundamental national security technology. The company warns that frontier AI systems could reshape cyber warfare, military planning, and mass surveillance. It claims the Chinese government already uses AI for censorship, monitoring, and cyber operations, and that more advanced systems could dramatically amplify these capabilities. The report also highlights concerns about safety practices at Chinese AI labs, noting that only a small number publicly disclose safety testing results for high-risk areas such as chemical, biological, radiological, and nuclear threats.
The paper argues that a closely contested AI race could weaken safety standards globally, as companies and governments feel pressured to release increasingly capable systems faster. Anthropic calls for a proactive approach: tightening controls on advanced chips, combating chip smuggling, and restricting overseas compute access used by Chinese firms. The company also urges the US to expand efforts promoting American AI systems worldwide to lock in a permanent lead.
Window of Opportunity Closing
The policy paper arrives as debate intensifies in Washington over semiconductor export restrictions and the long-term US AI strategy. Key questions include how much access Chinese firms should have to advanced Nvidia hardware. Anthropic states that the US currently holds a rare opportunity. "There is a high likelihood that we will look back on 2026 as the breakaway opportunity for American AI," the company wrote. At the same time, Anthropic supports AI safety dialogue with Chinese experts where possible and stresses that its concerns are directed at the Chinese Communist Party, not the Chinese people or the broader AI research community.
The background of the US-China technology rivalry stretches back several years. In 2022, the Biden administration imposed initial export controls on advanced AI chips, and those rules have been tightened multiple times since. The Trump administration has continued and in some areas intensified those restrictions, while also pursuing diplomatic engagements. The semiconductor industry is a critical battleground. Companies like Nvidia, AMD, and Intel have developed lower-performance chips specifically for the Chinese market to comply with export rules, but concerns persist about smuggling and unofficial channels. The AI race also encompasses talent migration, with many top Chinese researchers trained in US universities now returning to lead AI labs in China.
Anthropic's report emphasizes that the technology itself is dual-use: it can be a force for economic growth and scientific discovery or a tool for authoritarian control. The company's recommendations include strengthening the supply chain for advanced chips, increasing investment in domestic semiconductor fabrication, and enforcing international agreements on AI safety standards. The paper also calls for partnerships with allied nations to ensure they do not become backdoor suppliers of restricted technology.
Critics of export controls argue that they may ultimately harm US companies by limiting their market access and accelerating China's push for self-sufficiency. However, Anthropic dismisses this concern, arguing that the stakes of losing the AI race are far higher than short-term commercial losses. The company points to China's massive investment in AI research, its vast data resources, and its centralized government decision-making as factors that could allow it to leapfrog ahead if US controls are not watertight.
The report also raises the issue of algorithmic transparency. Anthropic notes that Chinese AI models are often opaque in their training and deployment, making it difficult to assess their safety or bias. By contrast, US companies publish more information about their systems, though the level of disclosure varies. The paper suggests that global norms for AI safety should be established in a way that incentivizes openness and penalizes opaque practices.
In summary, Anthropic positions 2026 as the critical inflection point. The decisions made in the next few months could determine which nation leads AI for the rest of the decade and beyond. The company’s analysis is a call to action for US policymakers to act decisively on export controls, enforcement, and international cooperation. Whether Anthropic’s scenario proves prescient or self-interested, the core message is clear: the window for securing democratic AI leadership is narrowing fast.
Source: eWEEK News