Is an AI kill switch a real thing?

California Governor Gavin Newsom has issued an executive order aimed at strengthening AI safety regulations, including the potential implementation of an 'AI kill switch' for frontier models. The order directs an expert-led panel to develop safety guidance and explore emergency shut-off mechanisms to prevent AI systems from causing catastrophic harm. While the concept of a universal off-button has gained traction in legislative circles, including a bipartisan bill in Congress, experts remain divided. Critics, such as the Center for Democracy and Technology, argue that binary safety regulations like kill switches may be ineffective, easily circumvented, or prone to government abuse. Prominent figures like Geoffrey Hinton have also questioned the long-term efficacy of such measures against superintelligence. As the debate continues, the focus remains on whether a single technical fail-safe is a viable solution or merely an aspiration in the rapidly evolving landscape of AI development.