H.R. 10152 is a bill of the 119th Congress introduced in the House of Representatives on August 27, 2026 by Representative Gabe Evans (R-CO-8) and referred the same day to the Committee on Energy and Commerce, with a committee meeting listed for September 1, 2026 at 10:15 a.m. Its long title is "To direct the Secretary of Commerce to support the adoption and use of American open artificial intelligence models, and for other purposes." This page summarizes the Introduced-in-House text, the only version available (Source: congress.gov). The topic page is Open-Source AI Leadership Act (H.R. 10152, 119th Congress).
The bill runs to five short sections: short title, a support mandate at section 2, an assessment and reporting mandate at section 3, a rule of construction at section 4, and eleven definitions at section 5. It carries no appropriation, no penalty provision, and no obligation falling on a private party.
The two-model taxonomy
Section 5 builds the whole bill on a pair of mirrored definitions distinguished by the developer's nationality rather than by the model's capability, license terms, or risk profile.
| Term | Definition (§ 5) |
|---|---|
| Open artificial intelligence model | A model with weights publicly released for download or distribution, or distributed under an open license permitting use, modification, and redistribution of source code and model weights |
| Qualified open model | An open model developed and made available in interstate or foreign commerce by a U.S. person, and not developed or made available by a covered nation or an entity or individual subject to a covered nation's control |
| Foreign adversary model | An open model developed or made available in interstate or foreign commerce by a covered nation, or by an entity or individual subject to a covered nation's control |
| U.S. person | A person that is domiciled in the United States, headquartered in the United States, and organized under the laws of the United States (all three) |
| Covered nation | The meaning given in 10 U.S.C. § 4872(f), the list used in defense acquisition |
| Qualified foreign partner | A foreign country or foreign political and economic union that is not a covered nation |
"Artificial intelligence" takes the meaning in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401); an artificial intelligence model is "a software component of an information system that incorporates artificial intelligence and uses computational, statistical, or machine-learning techniques to produce outputs from a defined set of inputs." Agency takes the meaning in 5 U.S.C. § 551, and State includes the District of Columbia, each commonwealth, territory, or possession, and each federally recognized Indian Tribe.
Because the two categories turn on the developer's nationality and control, a model that meets the open-weights test but is developed by neither a U.S. person nor a covered-nation entity — a model from an allied jurisdiction, for instance — falls into neither category and is outside both the support mandate and the assessment mandate.
Support for qualified open models (§ 2)
Section 2 directs that the Secretary "shall take any of the following actions as appropriate" — a mandatory verb applied to a menu from which the Secretary selects. The six listed actions are: designate a single point of contact to coordinate effort between the Department of Commerce, relevant private-sector entities, and other stakeholders; review any relevant program or power of the Secretary and direct program heads to support adoption; identify any barrier to adoption and act to address it; enter into agreements with private-sector entities including startups, agencies, states, and qualified foreign partners; develop policy and recommendations to facilitate adoption; and identify evidence-based approaches, best practices, frameworks, benchmarks, or other methods to monitor the adoption and use of qualified open models and foreign adversary models.
Assessment of foreign adversary models (§ 3)
Section 3(a) requires the Secretary to identify, assess, and make publicly known, as appropriate, any risk related to the adoption and use of foreign adversary models, enumerating six categories:
- how such a model was trained, "including any risk related to a covered nation, or an entity or individual that is subject to the control of a covered nation, conducting any such training through any unauthorized use of an artificial intelligence model";
- the confidentiality, integrity, and accessibility of personal and proprietary information, including risk of access by a covered nation or an entity or individual under its control;
- the security and resiliency of any organization that uses or adopts such a model, and consequent supply-chain risk;
- model output, "including with respect to historical accuracy, scientific inquiry, and freedom of expression";
- the efficacy, "or lack thereof," of any feature intended to prevent misuse or exploitation, including any safety or security feature; and
- the development and use of chemical, biological, radiological, or nuclear capabilities, or other similar risk to national and economic security.
The first limb reaches distillation or other training conducted through unauthorized use of another model. The fourth reaches content and censorship characteristics of a model's outputs rather than security properties.
Section 3(b) requires a report published on a public-facing website not later than 18 months after enactment and annually thereafter, covering the identified risks plus an overview of the adoption and use of leading foreign adversary models, their cost, capability, and performance, and a comparison of that information against the corresponding information for qualified open models where available. Section 3(c) requires submission to the House Committee on Energy and Commerce and the Senate Committee on Commerce, Science, and Transportation within 30 days of publication. Section 3(d) sunsets the publication requirement 10 years after enactment.
Rule of construction (§ 4)
Section 4 states in a single sentence that nothing in the Act "may be construed to authorize the Secretary to ban, restrict, or otherwise make unavailable an open artificial intelligence model in interstate or foreign commerce." The bar is written without a nationality qualifier: it protects the availability of any open model, foreign adversary models included, and confines the bill's instruments to coordination, agreements, policy development, assessment, and publication.
Provenance
Full text of the Introduced-in-House (ih) version, retrieved from congress.gov — the Library of Congress's official publisher of federal bill text — on August 29, 2026 via the queue task INGEST-hr-10152-open-source-ai-leadership-act.md. Dublin Core metadata served with the page independently records bill number H.R.10152, creator Rep. Evans, and date 08/27/2026, matching the document header. Only one version of the bill exists. Site chrome was removed at capture; the bill text is complete and unaltered. A Nextgov/FCW account of August 28, 2026 independently named Rep. Gabe Evans and described a bill directing Commerce to support adoption of open AI models; that account did not carry the bill's short title, which the text supplies.
Relationships
- related: Open-Source AI Leadership Act (H.R. 10152, 119th Congress) — the topic page this text underlies.
- instance-of: Open-Source AI / Open-Weight Models — federal promotion of open-weight models as a policy instrument.
- related: Open-Weight Frontier Models — the statutory open-weights definition turns on public release of weights or an open license.
- related: Export Controls (AI) — addresses covered-nation models by assessment and publication rather than trade controls.
- related: US-China AI Competition: Different Races, Different Metrics — the qualified/adversary split maps onto the U.S.–China frame.
- related: AI Advertising Disclosure Act (H.R. 10146, 119th Congress) — introduced the same week and referred to the same committee.