美国矽谷最初将人工智慧描述为一种可能灭绝物种的威胁，但现已转向加速主义叙事，任何质疑者都被贴上 AI 毁灭的标签。这种气氛下，恐惧在全国形成了罕见的跨党共鸣：一个清楚的多数认为 AI 的害处超过好处，NBC 的最新民调甚至将 AI 的净负面评价排在 ICE 之下。民主党与共和党选民都不小比例反对新建资料中心。尽管如此，主要 AI 企业在特朗普支持下的游说力量仍不断削弱联邦监管，让「AI 不可阻挡」的话语在政治上更具传播力；这也在美国这个庆祝建国 250 年的共和国中引发忧虑。
从技术上看，限制 AI 速度仍有空间。国会曾两次否决特朗普在七月《美国梦法案》（big beautiful bill）与十二月国防预算中加入的「先行排除权」（pre-emption）条款；每次都插入两页文字，若通过将禁止各州规范 AI。随后特朗普在行政命令中绕开了国会反对，并将儿童安全责任转交父母，由此放松平台约束。称之为「普罗朋斯主义」并不准确，他的政策更像寡头倾向：AI 产业繁荣始终是核心目标。Anthropic 的 Mythos 版本中提出模型发布前需申请审核，但该机制是自愿性的；且审核者为商务部长 Howard Lutnick，相当于让特朗普的内阁监督自身产业。
在 AI 与中国议题上，行政部门近期暗示将与习近平推进 AI 合作，表面上似乎修正了「中美是零和竞赛」的既有共识；但这一转向更像是商业导向。Nvidia 的 Jensen Huang 追求向中国出售更高阶晶片，特朗普对台湾垄断高端半导体也抱持不满，华盛顿之外部供应安排可能被拿来作为谈判杠杆。科技业在国内资助更强硬的反规范路线，却在对外却游说美方放宽对华销售，两面作法日后可能爆出冲突。要处理真正的全球 AI 风险，需要更广泛的美中合作；但在股市仍高度依赖大型科技估值的现况下，特朗普任内推行严格监管的机率极低，蓝领美国人依旧不是其政治优先。
Silicon Valley once framed AI as a species-level threat, but it has shifted to an accelerationist narrative where dissent is dismissed as AI doomerism. In this environment, fear of AI has become a rare bipartisan consensus: a clear majority believes AI will do more harm than good, and a recent NBC poll placed AI’s net-negative rating below ICE. A significant share of both Democrats and Republicans opposes new data centers. Yet lobbying power from major AI firms—amplified by Trump—has repeatedly weakened federal constraints, making the rhetoric of “AI is unstoppable” politically potent even as this occurs in the U.S. 250th year of the republic.
Technically, AI can be slowed. Congress twice rejected Trump’s proposals in July’s “big beautiful bill” and the December defense bill that inserted a pre-emption clause to block state regulation. Each insertion was two pages. After lawmakers rejected them, Trump used an executive order to retain his preferred outcome and shifted child-safety responsibility to parents, limiting constraints on platforms. Calling him populist misses the point; he is more plutocratic. Industry growth remains the priority. Anthropic’s Mythos-era idea that new models should seek approval was voluntary, and placing review under Commerce Secretary Howard Lutnick is akin to asking Trump to police his own administration.
On China, the administration has recently signaled a prospective AI opening with Xi Jinping, appearing to depart from the old zero-sum U.S. frame, though commercial interests dominate. Nvidia CEO Jensen Huang wants to sell more advanced chips to China, while Trump is frustrated by Taiwan’s control of high-end semiconductor supply, creating room for trade leverage. The tech sector is playing both sides: at home it funds hard-line anti-regulation voices, while abroad it presses for U.S. AI sales to China. This contradiction may become visible. Yet any serious mitigation of systemic AI risk needs broader U.S.-China cooperation, which remains unlikely under an administration where stock-market dependence on Big Tech valuations leaves little incentive for strong regulation while in office.