Secure and Accountable Military AI Act (Gillibrand, June 2026)

The Secure and Accountable Military AI Act is a US Senate bill introduced by Sen. Kirsten Gillibrand (D-NY) on June 2, 2026. It would restrict the Defense Department's use of AI in operations and regulate its use in fully-autonomous weapons, domestic surveillance, and nuclear missions, while establishing a chain of human accountability for AI used in combat (Source: defenseone.com). The bill was introduced without cosponsors, and Gillibrand plans to offer its elements as amendments to the FY27 National Defense Authorization Act (NDAA) rather than advance it as standalone legislation.

Status and timeline

Gillibrand announced the bill on June 2, 2026; the bill text is dated May 29, 2026. She introduced the measure without cosponsors and intends to advance it as NDAA amendments rather than as standalone legislation, a common path for defense-policy riders. A parallel effort by Sen. Elise Slotkin (D-MI) would tuck a related AI-guardrails bill (S.4113, introduced earlier in 2026) into the same NDAA. The Senate Armed Services Committee was slated to mark up the FY27 defense bill the following week.

"High-consequence" designation

The bill would direct Defense Secretary Pete Hegseth to designate specified AI uses — nuclear missions, lethal targeting, domestic surveillance, and cyber — as "high consequence." Such uses would require written approval from an undersecretary or the Vice Chairman of the Joint Chiefs of Staff before deployment.

High-consequence AI uses would also trigger a 15-day advance notification to the congressional defense committees, or, in certain circumstances, notification within 48 hours after deployment. The Pentagon would have to make clear who the accountable human decision-maker is, or what accountability chain exists, for AI-guided technology used during high-consequence operations, keeping a human responsible rather than diffusing accountability into the system.

Contractor incident reporting

The bill reaches frontier-AI vendors. AI contractors would have to rapidly report certain incidents to the Pentagon, including theft of model weights and data poisoning, within 3 days for security breaches and 7 days for concerning model behavior (Source: gillibrand.senate.gov).

Context and reactions

The bill arrived the same day President Trump signed the AI cybersecurity executive order, whose framing — that AI should be "deployed rapidly to confront any and all threats" and not stifled by "overly burdensome regulation" — sits in tension with Gillibrand's guardrails approach, reflecting an executive-versus-legislative split over military-AI risk tolerance.

The bill also lands against the backdrop of DoD–Anthropic friction earlier in 2026 over the company's objections to certain military uses of its technology, and amid persistently low US public trust in AI, which analysts have flagged as a national-security concern in its own right. Bloomberg Economics' defense lead characterized the bill as codifying long-standing norms as private AI firms become "more and more enmeshed with the Pentagon," describing it as a potential check on the Department's embrace of AI.

Relationships

• related: Autonomous Weapons, AI and National Security, Meaningful Human Review, Defense / Military — AI Deployment, Chief Digital and AI Office (CDAO), Anthropic
• contradicts: EO — Promoting Advanced AI Innovation and Security (Trump, signed June 2, 2026)
• instance-of: Risk-Based AI Regulation

Provenance note: Built from Defense One reporting and the sponsor's own section-by-section summary (gillibrand.senate.gov). The full bill text PDF (gillibrand.senate.gov/wp-content/uploads/2026/06/20260529-SAM-AI-Bill.pdf) is the primary source and would warrant a sources/ page only if the bill advances substantively (e.g., adopted as an NDAA amendment). confidence: medium — fresh introduction, status fast-decay.