Proposed US law would make AI risk reporting a legal obligation | CSO Online

What We Know

Lawmakers have introduced legislation called the AI Incident Reporting Act that would impose a legal duty on developers of advanced AI models to report critical or major safety and security incidents. Reporting obligations are aimed at forcing disclosure of dangerous model behavior and security threats that current oversight gaps leave unreported.

Reporting would be made to the federal government (reports indicate the Commerce Department would receive notifications), and multiple outlets describe the bill as setting a clear timeline for reporting: one notice says developers would need to flag dangerous or critical incidents within seven days. Coverage frames the bill as an attempt to close oversight gaps by bringing incident visibility and security risks to regulators’ attention.