Protection Against Foreign Adversarial Artificial Intelligence Act of 2025
- Sponsor
- Sen. Cassidy, Bill [R-LA]
- Committees
- Commerce, Science, and Transportation Committee (primary)
- Last action
- May 7, 2025
Bottom line
The bill aims to bolster U.S. national security by restricting the use of specific foreign-affiliated AI in federal contracting and by requiring a comprehensive assessment of threats posed by AI from adversarial nations.
What it actually does
This bill prohibits federal contractors from using the DeepSeek AI application, or any successor service from its developer High Flyer, for any work related to a federal contract, though it includes a waiver for national security or research purposes. Additionally, it mandates that the Secretary of Commerce, in consultation with the Secretary of Defense, submit a detailed report within one year on national security threats posed by artificial intelligence platforms based in or affiliated with "countries of concern," covering aspects like censorship, propaganda, export control circumvention, and data privacy risks.
Proponents argue
Supporters argue that this bill is a critical and proactive measure to safeguard sensitive government data, protect national security interests, and prevent foreign adversaries from exploiting artificial intelligence technologies for espionage, intellectual property theft, or the spread of state-sponsored propaganda. They emphasize the urgent need to address the growing risks associated with AI developed in countries that may pose a threat to U.S. interests, ensuring that federal operations are not compromised by potentially malicious or compromised software.
Opponents contend
Critics might contend that singling out specific applications or companies could stifle innovation, create unnecessary bureaucratic hurdles for federal contractors, or lead to retaliatory measures from other nations. They may also argue that a blanket prohibition could limit access to potentially superior or more cost-effective AI tools, or that the rapid evolution of AI technology could quickly render specific bans obsolete, requiring constant legislative updates.
The bill is very short and straightforward, making it highly accessible for any informed reader to understand its core provisions quickly.
Section 2(a)
Prohibition on Use of DeepSeek by Federal Contractors
This section establishes a direct prohibition: no federal contractor with an active federal contract is allowed to use the DeepSeek artificial intelligence application, or any future application or service developed by its parent company High Flyer or any of its subsidiaries. This ban applies to any activity that fulfills, assists, executes, or otherwise supports a contract with a federal agency, aiming to prevent the use of potentially compromised foreign AI in government work.
Supporters argue
Supporters argue this prohibition is a vital national security measure to prevent potential data exfiltration, espionage, or manipulation of federal government work by foreign adversaries. They highlight the specific risks associated with AI platforms developed in countries of concern, where data handling practices and algorithmic integrity may be subject to foreign government influence, posing a direct threat to sensitive information and operations.
Critics contend
Critics might contend that a specific ban on one application could be arbitrary or quickly become outdated as the artificial intelligence landscape rapidly evolves. They may also argue that it could place an undue burden on contractors who might already be integrated with such tools, potentially increasing compliance costs, delaying project completion, or limiting access to competitive technologies.
Tradeoffs
This provision balances the imperative to protect national security and sensitive government data against potential impacts on federal contractors' operational flexibility, innovation adoption, and the competitive landscape for AI tools in the government sector. It prioritizes security over convenience or potential cost savings.
Section 2(b)
Waiver Authority for DeepSeek Prohibition
This section grants the Secretary of Commerce the authority to waive the prohibition on using the DeepSeek application, but only on a case-by-case basis and in consultation with the Secretary of Defense. A waiver can only be issued if the use of the application or service is explicitly required for the completion of a national security-related objective of a specific contract or for legitimate research purposes, providing a narrow exception to the general ban.
Section 3
Mandated Report on AI Threats from Countries of Concern
This section mandates that the Secretary of Commerce, in consultation with the Secretary of Defense, submit a detailed report to Congress within one year of the bill's enactment. The report must analyze the national security threats posed by artificial intelligence platforms, including large language models and generative AI, that are based in or affiliated with "countries of concern." The report's contents must specifically address censorship laws, potential use for state-sponsored propaganda, the impact of export control circumvention on AI development, and various privacy and data security threats to U.S. data.
This subsection defines the term "country of concern" by referencing the existing definition of "covered nation" found in section 4872(f) of title 10, United States Code.
Section 3(a)
Why it matters:This is standard legislative practice to avoid redundant definitions and ensure consistency with existing federal law. By referencing an established definition, the bill streamlines its text and aligns its scope with existing national security frameworks, rather than creating a new, potentially conflicting definition.
Case for: Using an existing definition ensures legal consistency across different statutes and avoids ambiguity regarding which nations are considered "countries of concern" for national security purposes. It leverages established legal frameworks, making the bill's scope clear to those familiar with relevant defense legislation.
Case against: For a casual reader or someone unfamiliar with Title 10 of the U.S. Code, this cross-reference makes it less transparent which specific countries are being targeted. It requires additional research to fully grasp the geopolitical scope of the mandated report, potentially obscuring the full implications without explicit listing.
Estimated impact: This definition clarifies the geopolitical scope of the mandated AI threat report, directly influencing which nations' AI platforms will be subject to U.S. government scrutiny and potentially impacting diplomatic and trade relations with those countries. It sets the boundaries for future policy recommendations.