Secure and Accountable Military AI Act of 2026
Sponsored By: Senator Gillibrand, Kirsten E. [D-NY]
Introduced
Summary
Limits on autonomous weapons and tighter military AI oversight would bar most autonomous lethal systems and require clear human responsibility, strict testing, and fast reporting for high-risk AI used by the Department of Defense.
Show full summary
- Companies building frontier AI models for DoD would face fast security reporting and contract rules. They must report thefts or data-poisoning incidents within 72 hours and material vulnerabilities or dangerous model behavior within 7 days, and provide follow-up updates and protected handling of sensitive information.
- The Secretary would have to designate high-consequence AI uses within 180 days and a senior DoD official would need to sign off before deployment. Deployments would have to meet standards like realistic adversarial testing, legal review, operator override authority, logging, and continuous monitoring, and any non-exempt autonomous weapon system would need congressional authorization that lasts up to 3 years.
- The bill would bar AI for selecting or launching nuclear weapons and restrict AI collection or analysis targeting United States persons for tracking, pattern-of-life, watchlists, or risk scoring unless authorized by law. It permits narrow defensive AI uses for cybersecurity or force protection with informed consent.
Your PRIA Score
Personalized for You
How does this bill affect your finances?
Sign up for a PRIA Policy Scan to see your personalized alignment score for this bill and every other piece of legislation we track. We analyze your financial profile against policy provisions to show you exactly what matters to your wallet.
Bill Overview
Analyzed Economic Effects
4 provisions identified: 2 benefits, 0 costs, 2 mixed.
Bans on certain military AI uses
If enacted, the bill would bar the Defense Department from developing, buying, or using autonomous weapon systems except in narrow, listed cases that meet strict conditions. Any exception would require Secretary certification and 30 days' notice to congressional defense committees showing legal review and testing. The bill would also ban using AI to select targets for nuclear weapons or to execute nuclear launches. It would bar DoD from using AI to collect or analyze information about a United States person in the United States, except when an activity is expressly authorized by federal law and has a documented legal basis.
Stricter rules for high-risk military AI
If enacted, the bill would require the Secretary of Defense to set a process within 180 days to label Department AI as "high-consequence." Labeled categories would include AI for nuclear targeting support, lethal-targeting support, cyber effects outside DoD systems, autonomous weapon systems, and some domestic person-based analysis. The bill would bar deployment of any high-consequence AI unless a senior official makes a written determination that minimum testing, legal review, documented mission limits, operator controls, logging, and monitoring are in place. The Secretary would notify the Armed Services Committees at least 15 days before first deployment, with only a narrow 48-hour national-security exception.
Congress approval for autonomous weapons
If enacted, the bill would let the Secretary ask Congress to approve specific autonomous systems that otherwise would be barred. Such requests must include a classified annex with military necessity, operational limits, Law of Armed Conflict legal certification, a risk plan, and a human-judgment framework. The bill would force fast congressional action: a joint resolution must be introduced within 30 days, committees act within 20 days, and the Senate debate would be limited to 10 hours with no amendments. Any authorization would expire three years after enactment unless the Secretary renews it with a compliance certification.
New AI incident rules for contractors
If enacted, the bill would require DoD "covered contractors" on new and existing AI development contracts to report certain AI incidents. Some incidents would need an initial report within 72 hours and other incidents within 7 days. Reports would have to say what happened, dates, affected systems, whether model weights or data were accessed or altered, containment and remediation steps, and other specified details. The Secretary would notify the congressional Armed Services Committees within 7 days, issue contractor guidance within 90 days, and update acquisition rules within 180 days.
Sponsors & CoSponsors
Sponsor
Gillibrand, Kirsten E. [D-NY]
NY • D
Cosponsors
There are no cosponsors for this bill.
Roll Call Votes
No roll call votes available for this bill.
View on Congress.gov