FRONTIER Act vs AI Kill Switch Act: Congress's emergency AI shutdown mechanisms compared
Two new bills give government power to shut down AI models—but with very different rules.
The FRONTIER Act (Frontier Risk Oversight, National Transparency, Independent Evaluation, and Reporting Act) is the broader AI governance framework. Its Section 8 gives the Secretary of Commerce authority to issue emergency orders suspending a frontier developer's development, deployment, or internal use of a frontier model when it poses an 'imminent catastrophic risk'—defined as 50+ deaths or $1B+ property damage from a single incident. Provisional orders last up to 45 days and can convert into renewable final orders of up to 90 days. Violators face $10M/day fines, with willful violations carrying criminal charges up to 10 years in prison. The Act also creates licensing for third-party verification organizations and preempts state AI laws.
The AI Kill Switch Act takes a different approach. It authorizes the Secretary of Homeland Security, acting through CISA's Director, to suspend deployment of frontier models—but only after a 'covered incident' has occurred. It cannot act preemptively, and it cannot suspend training. The bar for a covered incident is lower than 'imminent catastrophic risk,' but the Act includes no specified order duration. Penalties are steeper: $20M/day. Notably, the AI Kill Switch Act explicitly requires frontier developers to maintain technical ability to shut down their models and mandates that CISA verify compliance—something the FRONTIER Act does not require. The FRONTIER Act is the exclusive mechanism for pre-incident catastrophic risk suspension; the AI Kill Switch Act is not exclusive. Together, they represent two competing philosophies on AI emergency response: proactive executive action versus reactive, incident-driven intervention.
- FRONTIER Act: Commerce Secretary can suspend training, deployment, or internal use for 'imminent catastrophic risk'; orders last 45 days extendable to 90; $10M/day fines.
- AI Kill Switch Act: DHS/CISA can suspend deployment only after a covered incident; no time limit; $20M/day fines; mandates kill-switch capability and compliance verification.
- Key difference: FRONTIER Act is proactive (pre-incident), AI Kill Switch Act is reactive (post-incident), with the latter explicitly requiring shutdown-ready technical infrastructure.
Why It Matters
Two competing AI shutdown frameworks could shape how the U.S. government responds to catastrophic AI risks—proactively or after the fact.