Covered AI Prohibition Act
- Sponsor
- Rep. Bacon, Don [R-NE-2]
- Committees
- Armed Services Committee (primary)
- Last action
- Sep 17, 2026
Bottom line
The bill mandates the Department of Defense to create and implement a system for identifying and eliminating AI technology from specific companies, aiming to enhance national security by restricting potentially risky foreign or adversarial AI from defense systems.
What it actually does
This bill, titled the 'Covered AI Prohibition Act,' amends existing law to require the Secretary of Defense to issue Department of Defense-wide guidance within 30 days of the enactment of the National Defense Authorization Act for Fiscal Year 2027. This guidance will establish processes for identifying and removing artificial intelligence products developed by 'covered artificial intelligence companies' from all DoD systems and devices. It also adjusts the effective date of an existing prohibition related to these AI products, making it active 90 days after a specified date.
Proponents argue
Proponents argue that this bill is a critical step to safeguard national security by preventing potentially compromised or adversarial artificial intelligence from being integrated into sensitive Department of Defense systems. They contend that clear guidance and a firm prohibition are necessary to mitigate risks associated with foreign-developed AI, ensuring the integrity and reliability of defense technology. This proactive measure protects against espionage, sabotage, and data exfiltration, which could undermine military operations and intelligence capabilities.
Opponents contend
Opponents might argue that without a clear definition of 'covered artificial intelligence companies' within this specific bill, the scope of the prohibition remains ambiguous, potentially leading to overreach or unintended consequences for legitimate technology providers. They could also contend that a broad prohibition might stifle innovation or limit access to cutting-edge AI technologies, potentially putting the U.S. at a disadvantage if allied or neutral companies are inadvertently swept into the ban. Furthermore, the administrative burden of identifying and removing such AI across all DoD systems could be substantial and costly.
The bill text is very short and straightforward, making it quick to read. However, understanding its full implications requires consulting the referenced prior legislation.
Section 2, paragraph (1) amending Section 1532(a)(2) of the National Defense Authorization Act for Fiscal Year 2026.
Requirement for DoD-Wide Guidance on Covered AI Products
This provision mandates the Secretary of Defense to create and issue department-wide rules within 30 days of the next fiscal year's NDAA becoming law. These rules will define how to identify specific artificial intelligence companies (referred to as 'covered artificial intelligence companies') and outline the procedures for preventing their AI products from being used in, and removing them from, all Department of Defense systems and devices.
Supporters argue
Supporters argue this guidance is essential for national security, providing a clear framework to protect critical defense infrastructure from potentially malicious or compromised AI technologies. It ensures a consistent approach across all DoD components.
Critics contend
Critics might argue that developing and implementing such comprehensive guidance within a tight 30-day timeframe could be challenging, potentially leading to rushed or incomplete policies. They may also raise concerns about the criteria for defining 'covered artificial intelligence companies' and the potential for overreach or unintended exclusion of beneficial technologies.
Tradeoffs
The tension lies between the urgent need for security against adversarial AI and the practical challenges of rapidly developing and implementing broad, effective guidance without stifling innovation or inadvertently harming partnerships.
Section 2, paragraph (2) amending Section 1532(a)(3)(B) of the National Defense Authorization Act for Fiscal Year 2026.
Modification of AI Prohibition Effective Date
This provision alters the language of an existing prohibition concerning artificial intelligence from certain companies. It specifies that the prohibition will take effect 'on and after the date that is 90 days after the date on which' a certain condition (not fully detailed in this bill) is met. This clarifies or delays the enforcement timeline of an existing ban.
The bill's core operative effect—the identification and prohibition of 'covered artificial intelligence companies'—is entirely dependent on the definitions and scope established in a previous law, the National Defense Authorization Act for Fiscal Year 2026. Specifically, the definition of 'covered artificial intelligence companies' is not provided in this bill.
Section 2, referencing 'Section 1532 of the National Defense Authorization Act for Fiscal Year 2026 (10 U.S.C. 2224 note)'.
Why it matters:This is standard legislative practice for amending existing law, rather than an attempt to be evasive. However, it does make the bill less self-contained for a casual reader.
Case for: Supporters would argue that this is efficient legislative drafting, avoiding redundancy by building upon existing legal frameworks and definitions.
Case against: Critics might contend that this approach makes it difficult for the public and even policymakers to fully grasp the bill's implications without extensive legal research, potentially obscuring the true scope of the prohibition.
Estimated impact: The impact is foundational, as the entire scope of the bill's prohibition hinges on these external definitions.