AI Emergency Button Act
- Sponsor
- Sen. Kennedy, John [R-LA]
- Committees
- Commerce, Science, and Transportation Committee (primary)
- Last action
- Sep 16, 2026
Bottom line
The bill establishes a fundamental safety requirement for advanced AI systems by mandating a human-controlled shutdown mechanism to ensure human oversight and control.
What it actually does
This bill mandates that any entity developing or operating an advanced artificial intelligence system in the United States must incorporate a technical capability allowing a human operator to shut down the system. It further directs the Secretary of Homeland Security, in consultation with relevant federal agencies, to issue regulations to ensure compliance within 90 days of the Act's enactment.
Proponents argue
Proponents argue that this bill is a crucial, proactive step in establishing basic safety protocols for rapidly advancing AI technologies. They emphasize the importance of a 'kill switch' to prevent potential runaway scenarios, mitigate unforeseen risks, and maintain human control over increasingly powerful systems, thereby building public trust in AI's development and deployment.
Opponents contend
Opponents might contend that the bill is overly broad or premature, potentially stifling innovation by imposing a one-size-fits-all technical requirement on diverse AI systems. They could also argue that defining 'advanced artificial intelligence system' and implementing a universal shutdown mechanism across all AI applications might be technically challenging or impractical, potentially leading to unintended consequences or new vulnerabilities.
The bill is exceptionally short and clearly written, making it very accessible for any reader to understand its core intent quickly.
Section 2(a)
Definitions of 'Artificial Intelligence' and 'Covered Entity'
This section defines key terms for the bill's application. 'Artificial intelligence' is defined by referencing an existing federal statute, the National Artificial Intelligence Initiative Act of 2020. A 'covered entity' is specified as any entity that develops or operates an 'advanced artificial intelligence system' within the United States, thereby clarifying who is subject to the bill's requirements.
Supporters argue
Supporters argue that clear definitions are essential for the effective implementation and enforcement of the bill, ensuring that the appropriate entities and technologies are targeted for regulation.
Critics contend
Critics might argue that the reliance on an external definition for 'artificial intelligence' and the lack of a specific definition for 'advanced artificial intelligence system' within this bill could lead to ambiguity and challenges in its practical application.
Tradeoffs
The provision balances the desire for concise legislation by referencing existing definitions against the potential for ambiguity if the referenced or undefined terms are themselves complex or subject to varying interpretations.
Section 2(b)
Mandatory Human-Controlled Shutdown Mechanism
This is the core operative provision of the bill, requiring all covered entities to ensure that their advanced artificial intelligence systems possess a technical capability for a human operator to shut down the system. This mandate aims to provide a direct and reliable means of human intervention and control over AI operations in critical situations.
Section 2(c)
Regulations for Compliance
This section directs the Secretary of Homeland Security, in consultation with other relevant federal agencies, to promulgate regulations within 90 days of the bill's enactment. These regulations will provide the specific details and standards for how covered entities must comply with the requirement to include a human-controlled shutdown mechanism in their advanced AI systems.
Defines 'artificial intelligence' by referencing section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401).
Section 2(a)(1)
Why it matters:This is standard legislative practice to avoid redundancy and maintain consistency with existing law, rather than an attempt to obscure the provision.
Case for: Referencing existing definitions streamlines legislation, avoids redundant text, and ensures consistency across federal policy regarding AI definitions.
Case against: Requires external research for full comprehension, potentially introducing ambiguity if the referenced definition is itself complex or subject to interpretation, making the bill less immediately accessible.
Estimated impact: This definition determines the foundational scope of AI systems that could potentially be covered by the entire Act, influencing which technologies are subject to the shutdown mandate.
The core requirement applies to 'advanced artificial intelligence systems,' but this specific term is not defined within the bill, nor is it explicitly referenced to an external definition.
Section 2(b) (and implied by 2(a)(2))
Why it matters:This could be an intentional choice to allow regulatory agencies flexibility in defining the scope as technology evolves, or it could be a strategic move to avoid contentious debates over specific technical definitions during the legislative process. It is not necessarily evasive but shifts a critical definitional task to the regulatory phase.
Case for: Allows for regulatory flexibility to adapt to rapid technological changes in AI, preventing the law from becoming quickly outdated by specific technical definitions.
Case against: Creates significant ambiguity for covered entities regarding which systems are subject to the mandate until regulations are promulgated, potentially leading to uncertainty, delayed compliance, and inconsistent application.
Estimated impact: The interpretation and definition of 'advanced' will critically determine the practical reach and burden of the bill, affecting which AI systems are required to implement a shutdown mechanism.