Under Federal Bill, AI Companies Would Need a ‘Kill Switch’

A new bipartisan bill introduced in the U.S. The legislation proposes fines of up to $2 million per day for companies that fail to comply with such an order. The bill's sponsors have not yet released the full text, but they describe the measure as a way to address scenarios where an AI system poses an imminent threat to national security or public safety.

The term "goes rogue" has not been precisely defined in public statements about the bill, raising questions about the criteria that would trigger a shutdown order. The Department of Homeland Security would be responsible for assessing whether a model meets that threshold. The fines are structured to escalate daily, meaning a company could face accumulating penalties if it continues operating the model after being ordered to shut it down.

It is one of several proposals aimed at addressing risks from advanced AI systems.