Following high-profile AI-related hacking incidents, an AI kill switch has been proposed as a safeguard to shut down dangerous AI systems. Proposed legislation like the AI Kill Switch Act and California Governor Newsom's executive order reflect this regulatory focus. However, President Trump has opposed such measures, and no national regulation including a kill switch exists. Experts like Dylan Baker argue that a kill switch is a problematic metaphor because modern AI involves complex, distributed infrastructure beyond a single switch's control. The metaphor also obscures the crucial human decisions and oversight involved in managing AI systems.The kill switch metaphor is rooted in similar rhetoric to the "black box" metaphor. Both metaphors, classified as structural metaphors by Lakoff and Johnson, simplify complex AI systems through physical concepts. While the black box emphasizes restricted incomprehensibility, the kill switch focuses on regaining control, yet both cultivate confusion by obscuring AI's distributed infrastructure and human influence.This simplification makes effective AI regulation difficult by failing to account for the infrastructural and human aspects of advanced models. The kill switch metaphor's first problem is that it falsely implies a centralized global AI infrastructure. In reality, AI models rely on a distributed web of computers and data centers. The second problem is that it conceals the human decision-making involved in AI training and deployment. It positions human intervention as an afterthought, rather than an integral and ongoing process.For example, accounts of a "rogue" OpenAI model hacking Hugging Face overlooked the fact that OpenAI personnel disabled safety mechanisms and set impossible tasks. Thus, the kill switch metaphor is ill-suited for AI regulation because it fails to account for AI's distributed infrastructure and human influence. Effective national regulation of AI requires a more nuanced understanding than a simple on/off switch.