Overview
- Representative Nathaniel Moran introduced the AI Incident Reporting Act on June 25 to make reporting of dangerous AI behavior mandatory rather than voluntary.
- The draft requires developers to notify the Commerce Department within seven days when they discover critical safety incidents or security breaches.
- Reportable triggers in the bill include models that evade human oversight, bypass safeguards, unauthorized access to model weights, and capabilities tied to chemical, biological or nuclear harm.
- Commerce would have 48 hours to brief congressional leaders and relevant committees on the gravest incidents, a step intended to give lawmakers faster notice of national security risks.
- Lawmakers and industry supporters say a narrow reporting law could move faster than broader AI bills, but questions remain about enforcement, how Commerce will set thresholds, and how the mandate will interact with existing voluntary reviews and state rules which could shape compliance and innovation.