Deterring American AI Model Theft Act of 2026

Bill Number

H.R. 8283

Origin Chamber

House

Congress

119th Congress, Session 2

Policy Area

International Affairs

Status

Introduced

Latest Action

2026-04-22: Ordered to be Reported (Amended) by the Yeas and Nays: 43 - 0.

Last Updated

2026-06-09T08:06:28Z

AI-Generated Summary

Purpose

The Deterring American AI Model Theft Act of 2026 aims to protect U.S. national security, economic interests, and intellectual property by deterring foreign adversaries from conducting model extraction attacks—unauthorized efforts to steal key technical features (like model weights or architectures) from closed-source AI models owned by U.S. companies. These models are proprietary AI systems not publicly shared, accessible only through owner-controlled interfaces like APIs under strict terms.

Key Provisions

• Sense of Congress: Affirms U.S. closed-source AI models are vital for security and economy; distinguishes legitimate research (following terms of service) from illegal extraction attacks.
• Definitions:
• Closed-source AI model: Proprietary AI with unreleased technical details, governed by owner terms.
• Country of concern: China (including Hong Kong/Macau), Russia, and others designated by the Secretary of State.
• Entity of concern: Foreign persons/entities from these countries or conducting attacks.
• Model extraction attack: Unauthorized high-volume querying that bypasses controls to replicate/improve another AI model (purpose inferred from patterns like query volume or coordination).
• Fraudulent account network provider: Entities enabling access via fake accounts to evade restrictions (excludes tools for free expression).
• Assessment and Reporting (Sec. 4):
• Secretary of State (with interagency group) assesses attacks and providers within 180 days; includes methods, origins, government involvement, detection, impacts.
• Secretary of Commerce submits report to Congress (210 days, annual updates for 3 years); establishes voluntary info-sharing with AI owners.
• Creates public AI Model Extraction Attackers List (maintained by State Dept., up to 5 years).
• Publishes best practices for detection/prevention.
• Deterrence Measures (Sec. 5):
• Commerce considers adding attackers/providers to the Entity List (restricts U.S. exports/tech to listed entities).
• President may impose sanctions (via IEEPA—a law allowing asset blocks in emergencies) on identified entities.
• Exceptions for humanitarian aid, UN obligations, U.S. intelligence.
• Violations carry civil/criminal penalties under IEEPA.

Significant Changes to Existing Law

• Introduces AI-specific definitions and concepts (e.g., model extraction attacks) not previously codified.
• Mandates new assessments, public lists, and info-sharing mechanisms building on export control laws (e.g., Export Control Reform Act).
• Expands Entity List criteria and enables targeted IEEPA sanctions for AI theft, without creating entirely new authorities.

Potential Impacts

• Government Agencies: Increases workload for State, Commerce (BIS), and interagency groups (e.g., Operating Committee for Export Policy) via assessments, reports, lists, and coordination.
• U.S. Citizens/Companies: Protects AI owners' IP; enables voluntary info-sharing and best practices, potentially reducing theft risks and boosting competitiveness.
• International Relations: Targets China/Russia; promotes diplomacy with allies; sanctions could limit trade/tech flows, escalating tensions but deterring adversaries.

Main Stakeholders Affected

• U.S. closed-source AI model owners: Primary beneficiaries via protection and support.
• Entities of concern and fraudulent providers: Face assessments, Entity List addition, sanctions.
• U.S. Government: State, Commerce, Congress (oversight via specific committees).
• Allies/Partners: Involved in diplomatic strategy.
• AI researchers/users: Clarifies legitimate vs. illegal activities.

Notable Legal, Constitutional, or Political Implications

• Legal: Relies on established frameworks (IEEPA, export regs) for enforceability; inference of attack intent from patterns may invite challenges but includes exclusions for compliant research.
• Constitutional: Sanctions target foreign entities/property in U.S. control, aligning with due process precedents; humanitarian exceptions mitigate overreach.
• Political: Bipartisan sponsorship potential (House Foreign Affairs referral); signals U.S. priority on AI dominance amid great-power competition, without restricting domestic innovation.

This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.

Sponsor

Rep. Huizenga, Bill [R-MI-4]

Cosponsors (16)

Rep. Moolenaar, John R. [R-MI-2], Rep. Lawler, Michael [R-NY-17], Rep. Crawford, Eric A. "Rick" [R-AR-1], Rep. Kamlager-Dove, Sydney [D-CA-37], Rep. LaHood, Darin [R-IL-16], Rep. Cline, Ben [R-VA-6], Rep. Harrigan, Pat [R-NC-10], Rep. McCaul, Michael T. [R-TX-10], Rep. Moore, Barry [R-AL-1], Rep. Salazar, Maria Elvira [R-FL-27], Rep. James, John [R-MI-10], Rep. Miller, Max L. [R-OH-7], Rep. Mackenzie, Ryan [R-PA-7], Rep. Gottheimer, Josh [D-NJ-5], Rep. Begich, Nicholas J. [R-AK-At Large], Rep. Moran, Nathaniel [R-TX-1]

Recent Actions

• 2026-04-22: Ordered to be Reported (Amended) by the Yeas and Nays: 43 - 0.
• 2026-04-22: Committee Consideration and Mark-up Session Held
• 2026-04-15: Referred to the House Committee on Foreign Affairs.
• 2026-04-15: Introduced in House
• 2026-04-15: Introduced in House

Bill Versions

• Deterring American AI Model Theft Act of 2026 — issued 2026-04-15 — PDF (19 pages)