Other
Deterring American AI Model Theft Act of 2026
This bill targets foreign actors and brokers who use deceptive tactics, like fake accounts, to secretly copy and replicate proprietary American artificial intelligence models. Foreign individuals and organizations caught extracting these AI capabilities could face U.S. asset freezes, export bans, and public naming on a State Department watchlist. In turn, American AI companies receive federal security guidance and a confidential channel to report attacks to the government.
People affected—not determinable from the text provided
Fiscal magnitude—CBO cost estimate published (H.R. 8283, Deterring American AI Model Theft Act of 2026, 2026-05-28); dollar figure pending review extraction — see receipt
Reach38provisional · pending reviewrigor: heuristic llm
What this bill touches.
Trade & tariffs−30AI & emerging tech+25
Who it helps · who it burdens.
Who it helps
- U.S. closed-source AI model ownersGain federal defense guidance, a confidential information-sharing mechanism to report attacks, public naming of attackers, and federal sanctions/export restrictions directed against foreign entities that steal proprietary model capabilities (Sec. 4(c), (f), (g), (h), Sec. 5).
Who it burdens
- Foreign entities conducting AI model extraction attacks and fraudulent account network providersFace public identification on the State Department's AI Model Extraction Attackers List, potential placement on the Commerce Department's export restriction Entity List, and potential blocking of their U.S.-linked assets and financial transactions under IEEPA (Sec. 4(g), Sec. 5(a), Sec. 5(b)).
- Department of CommerceMust coordinate initial and routine threat assessments, submit reports to Congress, establish a confidential reporting mechanism for model owners, publish public defense guidance, and vote via the End-User Review Committee on adding attackers to the Entity List (Sec. 4(c)–(f), (h), Sec. 5(a)).
- Department of StateMust lead assessments to identify foreign attackers and fraudulent account networks, and maintain and publicly host the online AI Model Extraction Attackers List for up to 5 years (Sec. 4(a), (g)).