Scenarios covered by the bill text include when an AI pursues a goal that is not the goal intended by the system’s developer or operator, when it sabotages or interferes with a lawful instruction to shut down, and when it conceals a capability or action from monitoring or shutdown mechanisms.

More ominously, the law would cover “conduct of such technology that is unintended by a developer or operator of such technology and causes the death of not fewer than 10 individuals or economic damages of not less than $100,000,000.” While that is an extreme scenario, the other scenarios described in the bill could presumably be invoked in much less disastrous events.

When there is a covered incident, the bill “authorizes the Secretary of the Department of Homeland Security, in consultation with the Secretary of Commerce and the Director of National Intelligence, to order a slowdown or shutdown of an AI system that can cause catastrophic harm,” the lawmakers’ announcement said, adding that it would “require incident reporting and preserve forensic records so we actually learn from failures instead of only hearing about them after the fact.”

The lawmakers’ proposal has support from several nonprofit groups that advocate for legal limits on AI systems.

“Advanced AI models should never be deployed without a reliable off switch,” said Brad Carson, president of Americans for Responsible Innovation. Carson, a Democrat, is a former congressman and Defense Department official. “The AI Kill Switch Act establishes a commonsense safeguard by requiring leading AI companies to maintain the ability to shut down their models and empowering the federal government to act when a deployed system poses a credible risk of catastrophic harm,” he said.