Defense AI Reliability and Reporting Act
Sponsored By: Representative Jacobs, Sara [D-CA-51]
Introduced
Summary
A centralized Department of Defense AI incident and vulnerability reporting program would collect, track, and analyze AI failures and weaknesses to surface recurring risks and guide fixes across the military. It emphasizes non-punitive reporting and requires remediation or deployment limits for serious, Department-wide problems.
Show full summary
- Service members and operators: Covered incidents include AI-caused harm, loss of control, safety or mission degradation, and failures to follow operator commands. Department-wide incidents must have a documented corrective action plan and validation before continued operational use.
- DoD acquisition, testing, and program offices: A designated official would standardize reports, run trend analysis, and issue guidance and alerts to inform testing, procurement, cybersecurity, and deployment decisions. The Secretary must deliver annual unclassified reports to Congress from 2027 through 2031 that summarize reports, trends, corrective actions, and policy recommendations, with a possible classified annex.
- Armed Forces members, civilian employees, contractors, and subcontractors: The bill would create a protected disclosure process that encourages good-faith reporting, shields reporters from retaliation, and requires protections for proprietary and classified information.
Personalized for You
How does this bill affect your finances?
Personalize government policy and PRIA will tell you what this bill means for your household, plus every other piece of legislation we track. PRIA reads each provision against your financial profile to show you exactly what matters to your wallet.
Bill Overview
Analyzed Economic Effects
3 provisions identified: 2 benefits, 0 costs, 1 mixed.
Annual AI reports to Congress
If enacted, this would require the Secretary of Defense to send an annual unclassified report to congressional defense committees in each year 2027 through 2031. Reports could include a classified annex. Each report would list the number and category of AI incident and vulnerability reports, summarize trends and corrective actions, and give recommendations for testing, procurement, cybersecurity, or deployment policy. For any covered AI incident that caused the death or bodily harm of a service member, the report would describe the systems, date and time, cause assessment, operational consequences, and corrective actions taken.
DoD central AI safety program
If enacted, this would require the Secretary of Defense to create a centralized, Department-wide program to collect, track, analyze, and fix AI incidents and vulnerabilities. It would cover problems from development, testing, buying, fielding, or operating AI systems. The Secretary would name an official to standardize reports, run trend analysis, and issue guidance. The program would promote non-punitive reporting and protect sensitive, proprietary, and classified information. Service members, DoD civilians, contractors, and subcontractors at any tier would be allowed to make protected, good-faith disclosures.
New limits on AI operational use
If enacted, this would require prompt reporting and categorization of AI incidents and vulnerabilities into local, program-level, or Department-wide response levels. For program-level or Department-wide matters, the designated official would coordinate remediation, retesting, mitigation measures, and deployment restrictions. For Department-wide incidents, the Secretary would require a written corrective action plan and validation that fixes were implemented before the affected system could be used again. The program would also require timely access to logs, system data, and model information to support analysis and fixes.
Sponsors & CoSponsors
Sponsor
Jacobs, Sara [D-CA-51]
CA • D
Cosponsors
Rep. Moran, Nathaniel [R-TX-1]
TX • R
Sponsored 8/31/2026
Rep. Whitesides, George [D-CA-27]
CA • D
Sponsored 8/31/2026
Roll Call Votes
No roll call votes available for this bill.
View on Congress.gov