Federal executive agencies would remove AI tied to certain foreign entities from their technology and stop buying it, with exceptions and national security waivers.
Within 60 days after enactment, federal officials would issue standards requiring executive agencies to remove certain foreign-linked AI software from their technology. Exceptions would cover law enforcement, national security, and security research, and case-by-case national security waivers could apply.
Executive agency leaders could not buy certain foreign-linked AI applications or software that uses them. The federal budget director could grant case-by-case national security waivers after consulting other named officials.
The federal budget director, consulting the General Services administrator and federal cybersecurity, intelligence, and defense officials, would set standards requiring agencies to remove covered AI. Covered AI includes applications developed or provided by a “foreign entity of concern”—a category defined in another federal law—and software using its general-purpose AI models. The standards must follow federal information-security requirements and include exceptions for law enforcement, national security, and security research. Agencies must document steps to reduce risk whenever they allow an excepted use. The budget director could grant a case-by-case national security waiver after consulting the named officials. The federal budget director must develop the standards and guidelines no later than 60 days after enactment.
Executive agency heads could not buy a covered AI application, including software that uses one. The purchase ban would cover AI applications developed or provided by a “foreign entity of concern,” software using their general-purpose AI models, and software using a covered application. The federal budget director could grant a case-by-case national security waiver after consulting the named federal officials.
No hearings or votes scheduled.