Duty of Care

Chapter 1: "The Kill Switch Delusion"

The afternoon of September 12, 2024, was unseasonably warm in Washington, D.C., and the hearing room on the third floor of the Rayburn House Office Building was doing what it always does when the cameras are on and the topic is terrifying: it was performing seriousness. The Subcommittee on Cybersecurity, Information Technology, and Government Innovation had convened to mark up H.R. 8342, the Artificial Intelligence Kill Switch Act of 2024. The bill’s title alone was enough to guarantee a full press gallery. A “kill switch” for AI. The phrase conjured a red button under a glass case, a final failsafe, the last line of defense between humanity and the machines. The members of the subcommittee, nearly all of them lawyers or career politicians, spoke of it in those terms. They spoke of “break glass in case of emergency.” They spoke of “a circuit breaker for the digital age.” The bill’s chief sponsor, a junior representative from a swing district with a tech-heavy donor base, put it most succinctly in her opening statement: “When the algorithm becomes a threat, we need the ability to turn it off. That is the minimum standard of accountability.”

It was a terrible idea, and not because the goal was wrong. The goal was unequivocally right. A society that has allowed artificial intelligence to be woven into the management of its hospitals, its transit systems, its financial markets, its criminal justice pipelines, and its military targeting software should have a