BLADE Act
- Sponsor
- Sen. Hagerty, Bill [R-TN]
- Committees
- Banking, Housing, and Urban Affairs Committee (primary)
- Last action
- Aug 5, 2026
Bottom line
The BLADE Act establishes a framework for the U.S. government to detect, deter, and punish foreign adversaries attempting to steal proprietary U.S. artificial intelligence capabilities through unauthorized model extraction.
What it actually does
This bill, known as the BLADE Act, directs the Secretary of Commerce to identify foreign persons and countries engaged in "model extraction attacks" against U.S.-owned closed-source artificial intelligence models or those facilitating such attacks. It mandates the creation of an "AI Model Extraction Attackers List" and requires the imposition of sanctions, including adding identified entities to the Entity List and blocking their property and interests under the International Emergency Economic Powers Act (IEEPA).
Proponents argue
Supporters argue that the bill is crucial for safeguarding U.S. national security and economic competitiveness by protecting the intellectual property of advanced AI models developed by American entities. They contend that foreign adversaries use model extraction to bypass years of research and development, gaining a shortcut to advanced AI capabilities, which poses a direct threat to U.S. technological leadership.
Opponents contend
Critics might express concerns that the broad definitions of "model extraction attack" and "person of concern" could inadvertently stifle legitimate AI research or create a chilling effect on international collaboration. They may also argue that the enforcement mechanisms, while strong, could lead to unintended economic consequences or retaliatory measures from targeted countries, potentially escalating tensions in the global tech landscape.
The bill is relatively short and clearly structured, allowing an informed reader to grasp its core provisions within a reasonable timeframe, though the technical definitions require careful attention.
Section 2
Sense of Congress on AI Model Protection
This section outlines Congress's view that U.S.-owned closed-source AI models are vital for national security and economic interests. It states that unauthorized acquisition of these models' capabilities by "persons of concern" (foreign adversaries) through "model extraction attacks" is a threat. It emphasizes the need for the U.S. government to identify, punish, and deter such attacks, while clarifying that authorized model training practices are distinct and legitimate.
Supporters argue
Proponents argue this section clearly articulates the national security imperative behind the bill, setting the stage for the subsequent operative provisions and signaling a strong commitment to protecting U.S. AI innovation.
Critics contend
No specific opposition is typically raised against a Sense of Congress, as it primarily expresses legislative intent rather than enacting direct policy. However, those wary of the bill's scope might see it as a broad justification for potentially overreaching measures.
Tradeoffs
This section balances the need to protect proprietary AI with the recognition of legitimate AI research, attempting to draw a clear line between malicious extraction and authorized training practices.
Section 3
Key Definitions for AI Model Protection
This section defines critical terms used throughout the Act. Key definitions include "closed-source artificial intelligence model" (proprietary, access via API, governed by terms), "country of concern" (initially China, Russia, plus others designated by Commerce), "person of concern" (foreign persons linked to countries of concern or conducting attacks), and "model extraction attack" (unauthorized replication of capabilities by circumventing controls, using fraud, or violating specific terms of service). It also clarifies that authorized training is not an attack.
Section 4
Assessment of AI Model Extraction Attacks
This section mandates the Secretary of Commerce, through the Under Secretary for Industry and Security (BIS), to conduct an assessment within 180 days of enactment. The assessment must identify persons and countries of concern engaged in model extraction attacks or acting as fraudulent account network providers. It requires analysis of attack methods, economic/national security consequences, and a diplomatic strategy. The section also mandates public consultation with AI owners, an annual report to Congress, a routine assessment process, an information-sharing mechanism, and the creation of a public "AI Model Extraction Attackers List."
Section 5
Deterrence of AI Model Extraction Attacks through Sanctions
This section mandates that entities identified in the assessment (Section 4) as having conducted or facilitated model extraction attacks, or their subsidiaries, be added to the Entity List maintained by the Bureau of Industry and Security (BIS). Additionally, the President is required to block and prohibit all transactions in property and interests in property of these "persons of concern" under the International Emergency Economic Powers Act (IEEPA). Exceptions are provided for humanitarian assistance and U.S. intelligence/law enforcement activities, and the President may waive sanctions under specific conditions.
The definition of "country of concern" extends beyond the explicitly named People's Republic of China and the Russian Federation to include any foreign country listed in Country Group D:5 in Supplement No. 1 to part 740 of title 15, Code of Federal Regulations, as published on January 1, 2026, that is designated by the Secretary of Commerce.
Section 3(3)(C)(i)
Why it matters:This is standard legislative practice to avoid listing every specific country in the bill text itself, instead relying on existing regulatory frameworks for dynamic updates. However, it means the full scope is not immediately apparent to a casual reader.
Case for: Supporters would argue this provides necessary flexibility for the Commerce Department to adapt to evolving geopolitical threats without requiring new legislation for each country. It leverages existing, well-understood regulatory lists.
Case against: Critics might argue that this technical reference makes the bill's full implications less transparent to the public and potentially allows for executive discretion to expand the scope of concern without explicit legislative debate on each addition.
Estimated impact: Potentially expands the list of countries subject to scrutiny and potential sanctions beyond the two explicitly named, impacting a broader range of foreign entities.
For the purpose of defining a "model extraction attack," the bill explicitly states that the purpose of extraction (to replicate, develop, train, or improve another AI model) "may be inferred from the totality of circumstances." It lists examples such as volume, pattern, coordination, concentration on specific capabilities, use of multiple accounts, or correlation with another AI model's development timeline.
Section 3(7)(B)
Why it matters:This is likely intended to provide flexibility for enforcement agencies to address sophisticated and evolving attack methods that may not leave explicit evidence of intent. It's a common approach in cybercrime legislation.
Case for: Proponents would argue that this inference clause is vital for effective enforcement, as malicious actors often conceal their true intent. It allows authorities to use circumstantial evidence to identify and prosecute attacks that are otherwise hard to prove.
Case against: Critics might express concern that allowing inference of purpose from "totality of circumstances" could lead to subjective interpretations or unintended targeting of legitimate research activities that might exhibit similar patterns (e.g., high volume queries for benchmarking).
Estimated impact: Grants significant interpretive power to the Commerce Department in identifying model extraction attacks, potentially broadening the scope of what can be deemed an attack.