SCALE Act
- Sponsor
- Rep. Moolenaar, John R. [R-MI-2]
- Committees
- Intelligence (Permanent Select) Committee (primary)
- Last action
- Apr 15, 2026
Bottom line
The SCALE Act seeks to dynamically adjust U.S. export controls on AI hardware to foreign adversaries, linking restrictions directly to their indigenous production capabilities and overall AI hardware accumulation, with the goal of maintaining U.S. leadership in advanced artificial intelligence.
What it actually does
The bill, titled the "Semiconductor Controls Adjusted to Limit Exports Act" or "SCALE Act," requires the Secretary of Commerce, in coordination with the Director of National Intelligence, to establish a new, systematic process for setting annual export control standards for AI hardware to "countries of concern." This process involves developing objective performance metrics for AI hardware, assessing indigenous production capabilities of adversaries, and setting an upper limit for exports based on these assessments, generally aiming to prevent adversaries from achieving AI hardware parity with the U.S.
Proponents argue
Proponents argue that the bill establishes a more strategic, data-driven, and adaptable framework for export controls, ensuring that restrictions evolve with technological advancements and adversary capabilities. They believe it will prevent foreign adversaries from developing advanced AI capabilities that could pose national security threats, while also providing greater business stability for U.S. companies through a predictable, annually assessed process.
Opponents contend
Opponents might argue that the bill's dynamic nature could create uncertainty for businesses, potentially hindering innovation or making it difficult for U.S. companies to compete in global markets if export thresholds are too restrictive or frequently change. Concerns could also be raised about the feasibility and accuracy of continuously assessing adversary AI hardware capabilities and the potential for unintended economic consequences or retaliatory measures from targeted countries.
The bill is moderately complex due to its technical definitions and procedural requirements, requiring a careful read to understand its full implications, which is manageable within a reasonable timeframe.
Section 2
United States Policy on AI Hardware and Software
This section outlines seven core policy objectives for the United States regarding artificial intelligence (AI) hardware and software. These include leading in global AI diffusion, ensuring foreign adversaries rely on U.S. supply chains, promoting U.S. AI hardware, preventing backfill among allies, limiting adversary indigenous AI production, providing business stability through systematic export controls, and preventing adversaries from achieving AI hardware parity with the U.S.
Supporters argue
Supporters argue this policy statement provides clear strategic direction for U.S. export control and promotion efforts, aligning economic and national security goals to maintain U.S. leadership in AI.
Critics contend
Critics might argue that some objectives, like 'crowding out their domestic markets,' could be seen as overly aggressive or could provoke retaliatory measures, potentially harming U.S. economic interests in the long run.
Tradeoffs
Balances aggressive containment of adversary AI capabilities with the need for U.S. industry leadership and business stability.
Section 3(a)
Establishment of Objective Performance Metrics for AI Hardware
This provision requires the Secretary of Commerce and the Director of National Intelligence to establish and publicly release objective performance metrics within 180 days to measure the state of AI hardware in 'entities of concern.' These metrics must assess individual AI hardware capabilities, aggregate indigenous production relative to demand, and the total estimated amount of AI hardware accessible to entities of concern. Specific categories include total processing power, interconnect bandwidth, and memory capacity bandwidth.
Section 3(b)
Annual Assessment of Adversary AI Hardware and Capabilities
This provision mandates the Secretary of Commerce and the Director of National Intelligence to submit a report to the appropriate congressional committees within 90 days of each metric update, assessing the quantity and quality of AI hardware produced in each 'country of concern' using the established metrics. The report must detail indigenous production, imported hardware, remotely accessible hardware, and compare adversary capabilities to U.S. AI hardware capacity, including a technical assessment of the most advanced AI hardware designed by an entity of concern.
Section 3(c)
Establishment of Rolling Annual AI Export Control Thresholds
This provision requires the Secretary of Commerce to adopt an export control policy for AI hardware, including remote access, to 'entities of concern.' Annually, an upper limit for exports will be set based on the assessment findings. This upper limit will be a policy of denial for license applications, set at no more than 110% of the performance of the most capable AI hardware item that meets a significant indigenous production threshold (25% of estimated annual aggregate demand). If no indigenous item meets this threshold, all exports are denied. Additionally, a license is required for any export, with a presumption of denial if approval would cause the aggregate adversary AI hardware to exceed 5% of the aggregate U.S. AI hardware.
Section 3(e)
Certification Requirements for Export Licenses
This provision specifies five conditions that an applicant must certify to for any license application to export AI hardware to an 'entity of concern' that is reviewed on other than a presumption of denial basis. These include assurances that the export won't delay U.S. orders, that the aggregate processing performance exported to entities of concern doesn't exceed 50% of shipments to U.S. customers, that items are not for military/WMD end-use, that independent third-party testing confirms technical capabilities, and that infrastructure-as-a-service providers won't transfer model weights or provide remote access to prohibited parties.
Defines 'indigenous production' of AI hardware as hardware physically fabricated within the territory of the People's Republic of China AND designed by any 'entity of concern.' It explicitly excludes AI hardware designed by an entity of concern but fabricated outside of any country of concern.
Section 5(8)
Why it matters:While standard legislative practice to define terms, this specific definition is critical for how the export control thresholds are calculated in Section 3(c) and could be easily overlooked by a casual reader focusing on the primary policy statements.
Case for: Supporters argue this precise definition is crucial for accurately assessing an adversary's true self-sufficiency in AI hardware, focusing controls on their ability to both design and manufacture domestically, which is key to preventing technological parity.
Case against: Critics might argue that excluding hardware designed by an entity of concern but fabricated elsewhere could create loopholes if those entities can still access advanced fabrication capabilities outside their home country, potentially undermining the bill's intent to limit adversary AI hardware accumulation.
Estimated impact: Directly impacts the calculation of the 'indigenous production threshold' (25% of demand) in Section 3(c)(1)(C), which in turn dictates the upper limit for AI hardware exports. This definition is fundamental to the bill's core mechanism.
Defines 'country of concern' by listing specific nations (People's Republic of China, Cuba, Iran, Democratic People's Republic of Korea, Russian Federation) and includes a mechanism for the Secretary of State to designate additional foreign countries. Such designation requires the country to be listed in Country Group D:5 under the Export Administration Regulations and for notice of the designation to be published in the Federal Register.
Section 5(5)
Why it matters:While defining scope is standard, the inclusion of a flexible designation process for additional countries is a significant grant of executive authority that determines the full reach of the bill's restrictions. Its placement in definitions might obscure its dynamic nature.
Case for: Supporters argue that allowing the Secretary of State to designate additional countries provides necessary flexibility to adapt to evolving geopolitical threats and ensures the export controls remain relevant and responsive to new national security challenges.
Case against: Critics might express concern about the potential for political rather than purely technical designations, or the lack of explicit congressional oversight for such additions, which could impact U.S. businesses operating in those countries without direct legislative review.
Estimated impact: Determines the full scope of countries subject to the stringent AI hardware export controls, potentially expanding or contracting based on future designations by the Secretary of State, thus directly affecting U.S. foreign policy and economic relations.
Defines 'remote access' as purposeful, knowing, reckless, or negligent access to an item subject to U.S. jurisdiction under this Act by a foreign person through a network connection (including the internet or a cloud computing service) from a location other than where the item is physically located, if the Secretary determines that the use of the item could pose a serious risk to U.S. national security or foreign policy.
Section 5(9)
Why it matters:While definitions are typically placed at the end, this particular definition significantly expands the traditional scope of export controls beyond physical goods to encompass virtual access. Its technical phrasing might obscure its broad impact on the digital economy and national security.
Case for: Proponents argue this definition is critical for closing potential loopholes where adversaries could circumvent physical export controls by accessing advanced AI hardware through cloud computing or other remote services, thereby maintaining the integrity and effectiveness of the control regime in the digital age.
Case against: Critics might argue that regulating remote access is technically challenging to enforce and could inadvertently restrict legitimate international collaboration or research, potentially stifling innovation or creating significant compliance burdens for cloud providers and other digital service companies.
Estimated impact: Significantly expands the scope of export controls to cover virtual access to AI hardware, impacting cloud service providers, U.S. AI hardware owners, and foreign persons accessing AI hardware remotely. This could fundamentally alter how advanced computing resources are shared globally.