BLADE Act

Bill S. 5252, the BLADE Act, would direct the U.S. government to identify and penalize foreign actors who improperly extract the technical secrets of closed-source, American-owned AI models, such as model weights and architectures. It requires the Commerce Department to assess and report on which "persons of concern" have carried out or facilitated these "model extraction attacks," publish a public list of identified attackers, and issue best-practice guidance for detecting and preventing such attacks. Entities placed on this list would face export restrictions (addition to the Commerce Department's Entity List) and economic sanctions blocking their property and transactions in the United States, with exceptions for humanitarian aid and law enforcement/national security activities, and a presidential waiver option requiring notice to Congress.

The bill targets foreign persons and entities—particularly those based in or controlled by China, Russia, or other designated countries of concern—who circumvent access controls, use fraudulent credentials, or violate terms of service to copy AI model capabilities. It explicitly does not apply to legitimate AI research conducted within a model owner's authorized terms of service. U.S. companies that own closed-source AI models would be able to voluntarily and confidentially report attacks to Commerce, and would benefit from the assessment, sanctions, and information-sharing mechanisms the bill creates.

The bill was introduced in the Senate on August 5, 2026, by Sen. Bill Hagerty with several cosponsors, read twice, and referred to the Senate Committee on Banking, Housing, and Urban Affairs. It would need to be approved by that committee, then pass the full Senate and House, and be signed by the President before becoming law; no further action has occurred yet.

Plain-English summary generated by Bill100 AI from the official record. Always verify against the source below.