A fascinating dichotomy has arisen within the US AI sector: the architects of the 'vehicles' advocate for a measured pace, whereas the passengers are eager to accelerate. Executives from companies like Anthropic, OpenAI, and xAI have urged a temporary halt in the pursuit of large-scale models, to allocate time for comprehensive safety research and independent evaluations. They contend that as autonomous agents become more pervasive, AI will acquire system-level operational privileges, heightening the potential for失控 (loss of control). Furthermore, with AI now contributing to its own development, the pace of iteration has surpassed the progress in safety measures and regulatory frameworks. Nevertheless, amidst fierce commercial rivalry, no company is inclined to be the first to decelerate; rather, they aspire to institute third-party evaluations and unified safety benchmarks. US President Trump, concerned about ceding technological supremacy to China, opposes any regulatory-induced slowdown in AI advancement. Beneath these divergent perspectives lies the China factor: the White House is apprehensive about relinquishing its dominant position, while Silicon Valley is preoccupied with the unpredictable nature of technological boundaries. Looking forward, it is improbable that the US AI industry will come to a complete standstill. A more pragmatic strategy would involve imposing speed restrictions in high-risk domains and standardizing incident reporting procedures. On the global stage, a crisis communication channel might be established initially, but the efficacy of interim safety measures remains to be seen.