Secure and Accountable Military AI Act of 2026

Bill Number

S. 4656

Origin Chamber

Senate

Congress

119th Congress, Session 2

Policy Area

Armed Forces and National Security

Status

Introduced

Latest Action

2026-06-02: Read twice and referred to the Committee on Armed Services.

Last Updated

2026-06-17T20:04:00Z

AI-Generated Summary

Secure and Accountable Military AI Act of 2026 (S. 4656)

Purpose

This legislation establishes requirements for the secure, accountable, and limited use of artificial intelligence (AI) by the Department of Defense (DoD). It focuses on oversight of high-risk applications, mandatory human accountability in critical decisions, contractor reporting of security incidents, and restrictions on certain AI uses, particularly involving autonomous weapons and domestic surveillance.

Key Provisions

• Definitions: Establishes terms including "artificial intelligence," "autonomous weapon system" (a system that selects and engages targets without further operator input after activation), "frontier artificial intelligence model" (large-scale models designated for national security relevance), "high-consequence artificial intelligence application" (uses with elevated risks such as nuclear command, lethal targeting, or cyber operations), and "covered incident" (events like model weight theft, data poisoning, supply chain compromises, or undisclosed concerning behaviors).
• High-Consequence Oversight (Section 3): Requires the Secretary of Defense to designate high-consequence AI applications. Prohibits operational deployment without written approval from a senior official (Under Secretary level or higher) after testing, legal review, documentation of limitations, fallback procedures, monitoring, and training. Mandates congressional notification 15 days in advance (or 48 hours in extraordinary circumstances).
• Human Accountability (Section 4): Mandates that AI may support but not replace human judgment in decisions involving force, detention, or domestic analysis. Requires a clearly identified accountable human decision-maker or chain with authority to override or terminate systems.
• Contractor Incident Reporting (Section 5): Requires covered contractors (those providing frontier AI models) to include reporting clauses in contracts. Mandates reports to the DoD within 72 hours for certain incidents (e.g., theft or compromise) or 7 days for others (e.g., vulnerabilities or concerning behaviors), with supplemental updates. The Secretary must notify Congress within 7 days and protect sensitive information.
• Limitations on AI Uses (Section 6): Prohibits AI for selecting or executing nuclear weapons, certain domestic person-based analysis on U.S. persons (e.g., tracking, risk scoring, or creating dossiers without legal authorization), and most autonomous weapon systems. Allows exceptions for semi-autonomous systems, operator-supervised systems in local defense against time-critical attacks, missile interception, and non-lethal uses, subject to legal review, testing, human controls, and constraints. Additional rules bar uses based solely on protected First Amendment activity or to evade privacy protections.
• Implementation (Section 7): Directs updates to acquisition regulations and guidance within 180 days, plus an initial briefing to Congress.
• Authorization Process for Non-Excepted Systems (Section 8): Allows the Secretary to petition Congress for approval of autonomous weapon systems outside defined exceptions via a joint resolution, with specific requirements including military necessity justification, legal opinions on law of armed conflict compliance, and risk mitigation plans. Includes expedited congressional procedures.

Significant Changes to Existing Law

This bill introduces new statutory requirements for AI governance in the DoD, including mandatory senior-level approvals, contractor reporting obligations, and explicit prohibitions on certain autonomous weapon systems and domestic AI surveillance. It creates a congressional approval mechanism for non-excepted autonomous systems, which did not previously exist in this form. It also codifies human oversight policies and incident reporting that build upon but go beyond existing DoD directives.

Potential Impacts

• Government Agencies: Increases administrative and compliance burdens on the DoD, including testing, documentation, and inter-agency coordination (e.g., with the Chief Digital and Artificial Intelligence Officer and National Security Agency). May delay deployments of AI systems.
• Citizens: Strengthens privacy protections for U.S. persons by limiting AI-driven domestic surveillance and analysis.
• International Relations: Could shape global discussions on autonomous weapons by imposing strict limits and human judgment requirements, potentially affecting alliances or arms control talks.
• Contractors and Industry: Imposes new security reporting duties on AI developers, with potential effects on contract terms and innovation timelines.

Main Stakeholders Affected

• Department of Defense components and leadership.
• Contractors developing or providing frontier AI models and systems.
• Congressional Armed Services Committees.
• U.S. persons subject to potential domestic AI analysis.
• Military operators, commanders, and legal reviewers involved in AI deployment.

Notable Legal, Constitutional, or Political Implications

The bill emphasizes compliance with the law of war, treaties, and constitutional protections (e.g., First Amendment rights and limits on surveillance of U.S. persons). It requires legal reviews and human accountability in use-of-force decisions. Politically, it establishes a structured congressional role in approving certain autonomous systems while promoting transparency through notifications and briefings.

This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.

Sponsor

Sen. Gillibrand, Kirsten E. [D-NY]

Recent Actions

• 2026-06-02: Read twice and referred to the Committee on Armed Services.
• 2026-06-02: Introduced in Senate

Bill Versions

• Secure and Accountable Military AI Act of 2026 — issued 2026-06-02 — PDF (33 pages)