Open-Source AI Leadership Act
- Sponsor
- Rep. Evans, Gabe [R-CO-8]
- Committees
- Energy and Commerce Committee (primary)
- Last action
- Sep 16, 2026
Bottom line
The bill aims to bolster American leadership in open-source AI by promoting domestic models and scrutinizing those from geopolitical rivals, without imposing outright bans on any models.
What it actually does
This bill directs the Secretary of Commerce to actively promote the adoption and use of American-developed open artificial intelligence models, referred to as 'qualified open models.' Concurrently, it mandates the Secretary to identify, assess, and publicly report on risks associated with 'foreign adversary models,' which are open AI models developed by covered nations or entities under their control. The bill explicitly states that it does not authorize the Secretary to ban or restrict any open AI model.
Proponents argue
Supporters argue this bill is crucial for national security and economic competitiveness, ensuring that the United States maintains a leading edge in the rapidly evolving field of artificial intelligence. Promoting American-developed open AI models fosters innovation, transparency, and trust, while assessing foreign adversary models helps mitigate potential risks like data exfiltration, supply chain vulnerabilities, and misuse for nefarious purposes. It is seen as a proactive step to secure the AI ecosystem.
Opponents contend
Critics might argue that focusing on 'American' models could lead to protectionist policies that stifle global collaboration and innovation in open-source AI, which thrives on international contributions. They might also contend that the assessment of 'foreign adversary models' could be perceived as discriminatory, potentially hindering academic research or commercial partnerships with entities that happen to be based in or associated with 'covered nations,' even if their models are technically sound and openly available. Some might also argue that the bill's scope is too narrow, focusing only on 'open' models and potentially overlooking risks from closed-source foreign models.
The bill is concise and can be read quickly, allowing ample time for informed evaluation.
Section 2
Support for Adoption and Use of Qualified Open Models
This section directs the Secretary of Commerce to actively support the use of American-developed open artificial intelligence models. This includes designating a point of contact, reviewing existing programs, identifying and addressing barriers to adoption, entering into agreements with private sector entities, agencies, states, and qualified foreign partners, and developing policy recommendations. The goal is to facilitate the widespread use of these models in commerce.
Supporters argue
Supporters argue this provision is essential for fostering a robust domestic open AI ecosystem, ensuring American leadership, and providing secure, transparent AI solutions for U.S. entities. It leverages government resources to accelerate innovation and adoption.
Critics contend
Critics might argue that direct government intervention in promoting specific models, even if 'American,' could distort the market or lead to less optimal solutions if superior foreign models are overlooked due to national origin.
Tradeoffs
Balances the desire to promote domestic innovation and security with the principles of a free market and global collaboration in open-source development.
Section 3(a)
Identification and Assessment of Risks from Foreign Adversary Models
This provision mandates the Secretary of Commerce to identify, assess, and publicly disclose risks associated with the use of 'foreign adversary models.' These risks include unauthorized training, data confidentiality and integrity, supply chain security, output accuracy and freedom of expression, efficacy of safety features, and potential for developing weapons of mass destruction.
Section 3(b)
Annual Public Report on Foreign Adversary Models
The Secretary is required to publish an annual report, starting 18 months after enactment, assessing risks from foreign adversary models. This report must include an overview of their adoption, cost, capability, and performance, and compare this information with available data on qualified open models. This reporting requirement sunsets after 10 years.
Section 4
Rule of Construction (No Ban or Restriction Authority)
This section explicitly states that nothing in the Act should be interpreted as authorizing the Secretary of Commerce to ban, restrict, or otherwise make unavailable any open artificial intelligence model in interstate or foreign commerce. This clarifies that the bill's intent is to promote and assess, not to prohibit.
Section 5
Definitions
This section defines key terms used throughout the Act, such as 'agency,' 'artificial intelligence,' 'artificial intelligence model,' 'covered nation,' 'foreign adversary model,' 'open artificial intelligence model,' 'qualified foreign partner,' 'qualified open model,' 'Secretary,' 'State,' and 'U.S. person.' These definitions are critical for understanding the scope and application of the bill's directives.
The bill defines 'foreign adversary model' as an open AI model developed or made available by a 'covered nation' or an entity/individual controlled by one. Conversely, a 'qualified open model' is defined as an open AI model developed and made available by a 'U.S. person' and *not* by a 'covered nation' or controlled entity. The definition of 'covered nation' refers to Section 4872(f) of title 10, United States Code, and 'U.S. person' has specific domicile, headquarters, and organization requirements.
Section 5(5) and 5(8) (Definitions of 'Foreign Adversary Model' and 'Qualified Open Model')
Why it matters:It is standard legislative practice to place definitions at the end, but the reliance on external code sections for 'covered nation' requires additional research to fully grasp the scope.
Case for: Supporters argue that precise definitions are necessary to clearly delineate which models are to be promoted and which are to be scrutinized, ensuring the bill's objectives are met without ambiguity. Referencing existing law for 'covered nation' provides consistency with broader U.S. policy.
Case against: Critics might argue that the definitions, particularly 'U.S. person' and 'covered nation,' could be overly restrictive or broad, potentially excluding valuable international collaborations or inadvertently targeting entities not intended by the spirit of the law. The reliance on external definitions can make the bill harder to interpret for a casual reader.
Estimated impact: High, as these definitions determine the entire scope of the bill's application and the entities it targets for promotion or assessment.