With the midterm elections behind it, Congress faces a familiar obligation: negotiating and passing the National Defense Authorization Act (NDAA) for fiscal year 2027. But this year’s defense bill carries implications that extend well beyond military spending. It could help determine whether the United States retains its advantage in artificial intelligence or inadvertently accelerates China’s technological ambitions.

As House and Senate leaders reconcile their competing priorities, one objective should take precedence: ensuring that the United States, rather than China, leads AI innovation and supplies the technology on which much of the world depends. Partisan gridlock has paralyzed much of Washington, but confronting China’s technological ambitions remains a rare area of bipartisan agreement. Lawmakers should judge every proposed restriction against that objective. Three measures under consideration for the FY2027 NDAA—the AI OVERWATCH Act, the Chip Security Act, and the MATCH Act—risk failing that test.

Their intentions are understandable. American technology should not strengthen China’s military, and sensitive capabilities must be protected. But national security involves more than denying adversaries access to individual products. It also depends on preserving America’s industrial strength, maintaining relationships with allies, and sustaining the technological influence that underpins U.S. leadership.

Washington needs an approach that distinguishes genuinely dangerous technology transfers from legitimate commerce that strengthens American companies and keeps international customers invested in U.S. technology. These proposals threaten to blur that distinction.

Consider the AI OVERWATCH Act. The House proposal would give Congress a direct role in certain AI chip export decisions, establishing a 30-day review period during which lawmakers could block a transaction through a joint resolution. The Senate’s approach differs, making careful scrutiny of the final NDAA language particularly important.

Congress unquestionably has a legitimate oversight role. But oversight should not become day-to-day political management of a rapidly evolving commercial market. Export decisions require technical expertise, intelligence assessments, and the flexibility to respond to changing technologies and strategic circumstances. Those responsibilities belong primarily to the executive branch for good reason.

Introducing another layer of congressional politics into export licensing risks making American suppliers less reliable without necessarily diminishing China’s technological capabilities. Customers investing billions in computing infrastructure need confidence that the equipment they purchase will actually arrive. If American suppliers become unpredictable, those customers will seek alternatives. America’s competitors will be more than willing to accommodate them.

The strategic choice should be obvious. Does the United States benefit more when customers worldwide depend on American hardware, software, and technical standards, or when they invest in competing ecosystems? Congress should protect those relationships rather than jeopardize them through the AI OVERWATCH Act.

The Chip Security Act presents a different danger: transforming a legitimate enforcement objective into an expensive and potentially unworkable technological mandate.

The Semiconductor Industry Association has warned that the proposed security mechanisms remain untested and could prove impractical on a global scale. Its concerns deserve attention. Location verification may eventually become an effective tool for enforcing export controls, but Congress should not confuse technological promise with demonstrated reliability.

Research into chip verification methods has identified significant technical limitations, including the possibility of evasion. No single mechanism offers a foolproof solution. Implementation also presents serious challenges. New communication and authentication functions require extensive cybersecurity testing to ensure they do not introduce vulnerabilities into the very systems they are intended to protect.

Engineering realities, rather than legislative impatience, should determine compliance timelines.

The MATCH Act raises an equally troubling concern: it threatens to undermine the international partnerships on which effective export controls depend.

The diplomatic consequences are already becoming apparent. The Netherlands, a critical player in the global semiconductor industry, has formally objected to the legislation’s extraterritorial reach. Dutch officials have warned that its sweeping provisions could damage domestic semiconductor companies and interfere with maintenance and technical support for equipment already delivered to customers.

Dutch Trade Minister Sjoerd Sjoerdsma has carried those concerns directly to Washington, warning against an approach that risks turning cooperation among allies into coercion.

Japan has also reportedly raised objections. Yet Representative Michael Baumgartner, a leading advocate of the legislation, has continued pressing for its passage. Japan is among America’s most consequential strategic partners in the Indo-Pacific, and its semiconductor industry is indispensable to the global technology supply chain. When Tokyo raises concerns about American trade restrictions, Washington should listen rather than dismiss them.

Congress must also consider the economic consequences that extend beyond diplomatic friction. The Center for Strategic and International Studies has warned that the MATCH Act could disrupt the global memory chip supply chain, potentially slowing the deployment of American AI systems. That would undermine the very technological leadership these restrictions are intended to preserve.

All three proposals suffer from a similar weakness: they risk measuring national security resolve by the number of new restrictions imposed rather than by whether those restrictions actually strengthen America’s competitive position.

Maintaining U.S. leadership in artificial intelligence requires a coherent strategy, not a collection of poorly coordinated legislative mandates.

Washington should protect sensitive technologies, target military end users and illicit procurement networks, and equip enforcement agencies with the personnel, investigative capabilities, and intelligence resources necessary to stop illegal diversions. It should negotiate practical export controls with allies and rigorously test verification technologies before imposing sweeping compliance requirements.

Meanwhile, Beijing is hardly waiting for Washington to finish its paperwork.

In September, Huawei unveiled its next-generation AI chip roadmap, while DeepSeek announced software tools designed to challenge Nvidia’s technological ecosystem. Restrictions that have limited American companies’ access to the Chinese market have also created incentives for Chinese firms to develop their own alternatives.

That should give Congress pause. Export controls may be necessary to protect national security, but poorly designed restrictions can produce unintended consequences, including encouraging the emergence of competitors that American companies might otherwise have kept at bay.

Before incorporating the AI OVERWATCH Act, the Chip Security Act, or the MATCH Act into the FY2027 NDAA, lawmakers should reconsider whether these measures advance American technological leadership or merely create new obstacles to sustaining it.

The purpose of America’s defense legislation should be to make it harder for China to close the technological gap, not harder for the United States to maintain its lead.

Winning the AI race requires more than restricting China’s access to American technology. It requires ensuring that the rest of the world still wants to use it.