Congress May Accidentally Help China Win the AI Race

C hina’s recent progress in artificial intelligence and high-performance computing should not cause panic in Washington, but it should remind Congress that America’s technological lead is no longer so overwhelming that it can afford self-inflicted policy mistakes.

That is why lawmakers should proceed carefully as they consider the Remote Access Security Act (RASA). The legislation is trying to solve a real national security problem. But if Congress writes it, or the executive implements it too broadly, the law could make American AI infrastructure harder for allies and neutral countries to use, chilling the growth of our technology sector and creating an opening for Chinese competitors.

That concern has become more urgent considering China’s recent technological achievements. Last month, Tech Digest reported that China’s LineShine system claimed the top spot in the global supercomputer rankings, powered by domestically designed processors. The Associated Press similarly noted that the system marks China’s first return to the top of that list since 2017.

That is a significant development, but it should not be mistaken for proof that China has surpassed the United States in AI.

Reuters noted that, while China reclaimed the top spot on the Top500 list, the biannual ranking of the world’s most powerful supercomputers, this benchmark is not primarily designed for AI workloads. That distinction matters because LineShine ranked lower on a benchmark that more closely reflects AI performance.

In other words, China’s supercomputing gains are real, but the broader strategic lesson is not that America has lost the AI race. China is steadily reducing its dependence on U.S. technology and becoming more capable of competing in the infrastructure that will support future AI systems. Washington should take that trend seriously and avoid policies that unnecessarily strengthen China’s competitive position.

The basic problem behind RASA is straightforward: The federal government imposes export controls to prevent advanced AI chips, supercomputers, and other sensitive U.S. technologies from reaching adversaries such as China, Russia, Iran, and North Korea. But a foreign actor may not need to buy restricted chips if it can rent access to the same computing power through cloud services.

RASA aims to close that loophole. The legislation would give the Commerce Department’s Bureau of Industry and Security (BIS) authority to regulate who may remotely access certain U.S. technologies, even if the hardware remains in the country.

The objective behind RASA is entirely legitimate. The Carnegie Endowment for International Peace has described cloud controls as a tool that could, under some circumstances, be used to curtail China’s AI capacity. But Congress should be careful not to cast the net so broadly that it catches America’s allies along with its enemies.

The risk is that legitimate foreign companies may face uncertainty, compliance costs, or legal exposure when using U.S. cloud infrastructure. This would raise the cost of doing business with U.S. AI providers, weakening their competitive position.

Foreign customers provide billions in cloud revenue and help reinforce the network effects that keep American AI companies dominant. Yet, depending on how RASA is interpreted and implemented, companies that employ Chinese nationals or multinational engineering teams could be unsure whether they are legally permitted to use American AI services, even if they have no connection to China’s government or military.

Consider a manufacturer in Germany, a software firm in the United Kingdom, or an AI startup in Singapore. A company with no military ties may still decide that building on American cloud infrastructure is too risky if the compliance rules are unclear. Some may turn to European, domestic, or regional alternatives. Others may look to Chinese providers such as Huawei, Alibaba, Tencent, or newer AI players such as DeepSeek.

Every customer America pushes away creates an opening for Chinese competitors to expand their commercial footprint, shape global technology standards, and deepen their geopolitical influence. That outcome is not guaranteed, but it presents a serious strategic risk.

This concern is not limited to speculation. The Center for Strategic and International Studies (CSIS), in its work on cloud competition, has warned that China supports its companies in gaining market share as part of a broader effort to increase its political influence.

In the same report, CSIS argued that America has an opportunity to shape global cloud infrastructure by making trustworthy U.S. and allied cloud services attractive to foreign customers. That argument cuts both ways. If U.S. rules become too burdensome or unpredictable, Washington could weaken one of its strongest advantages. Computing is a competitive global market, and customers — both foreign and domestic — care about cost, reliability, legal certainty, and speed.

This does not mean America should give up on export controls. Policymakers should continue to deny advanced AI capabilities to adversarial militaries, intelligence services, and state-backed actors. But the better approach is to target the threat more precisely.

Congress should make clear that the goal is to restrict hostile access to advanced AI compute, not to discourage legitimate companies in allied or neutral countries from building on American platforms.

An improved version of RASA would include clearer rules for trusted foreign customers, reasonable compliance pathways for multinational firms, and explicit guardrails to avoid penalizing legitimate commercial users with no connection to adversarial governments or militaries.

U.S. export controls on AI chips have already prompted China to accelerate the development of domestic alternatives. That does not mean export controls were a mistake. It means Congress should understand the trade-off. Restrictions can slow adversaries, but they can also motivate competitors, create new markets for non-U.S. suppliers, and push foreign customers to diversify away from American technology. In the context of cloud computing, that trade-off may be even more sensitive because customers can switch providers without waiting for the physical supply chain to change.

The United States can block adversarial military access to advanced AI compute while preserving trusted global access to American platforms. This balance is what will keep the United States ahead. If Congress gets the balance wrong, it may accidentally help China win customers, influence, and eventually a larger share of the world’s AI future.