Experts Advocate for Human-Controlled 'Kill Switches' in Advanced AI Systems
Bernie Sanders, Steve Case, and Susan Rice, among other experts, are advocating for the implementation of human-controlled, independently enforceable 'kill switches' in all consequential Artificial Intelligence (AI) systems. This proposal emphasizes that no AI system should operate beyond meaningful human intervention. Microsoft has also supported this concept since 2023, promoting 'safety brakes' to maintain human control over advanced AI. The experts argue that AI systems controlling critical infrastructure and autonomous functions should be run within secure cloud environments, incorporating layered safeguards that allow for intervention, correction, and shutdown. They draw a parallel to Elisha Otis's 1850s safety brake for elevators, which built public trust and enabled modern cities, suggesting a similar principle is needed for AI to ensure public confidence and progress.
