The United States is experiencing a period of institutional tension characterized by conflicting AI governance strategies between the federal administration and state-level leadership, specifically within California.

美國目前正經歷一段制度緊張期，其特點是聯邦政府與州級領導（特別是加州）在 AI 治理策略上存在衝突。

The current geopolitical landscape is marked by a struggle for primacy between public authority and private innovation. The federal administration has increasingly utilized export controls and security designations to constrain frontier AI models. This is exemplified by the restriction of Anthropic's Fable 5 model due to national security concerns and the subsequent designation of the firm as a supply chain risk by the Pentagon. Such measures follow a period of policy volatility, shifting from a laissez-faire approach to one of stringent oversight, partly precipitated by the perceived capabilities of the Mythos model to identify software vulnerabilities.

目前的地緣政治格局是以公共權力與私人創新之間的主導權爭奪為特徵。聯邦政府日益利用出口管制和安全界定來限制前沿 AI 模型。例如，由於國家安全考量，Anthropic 的 Fable 5 模型受到限制，隨後五角大廈將該公司定為供應鏈風險。此類措施是在一段政策波動期後採取的，從自由放任轉向嚴格監督，部分原因是發現 Mythos 模型具有識別軟體漏洞的能力。

Parallel to these federal constraints, the State of California has pursued a policy of rapprochement with the private sector. Governor Gavin Newsom has formalized an agreement with Anthropic to integrate the Claude AI system across all state and local agencies at a fifty percent discount. This initiative, supported by a March executive order, seeks to decouple state procurement from federal designations, thereby facilitating government efficiency while the administration continues to clash with the provider over the deployment of autonomous weaponry and mass surveillance.

與這些聯邦限制平行，加州政府則採取了與私營部門趨向和解的政策。州長 Gavin Newsom 已與 Anthropic 正式達成協議，以五折優惠將 Claude AI 系統整合至所有州和地方機構。這項倡議由三月的行政命令支持，旨在將州政府採購與聯邦界定脫鉤，從而提升政府效率，而聯邦政府則繼續就部署自主武器和大規模監控與供應商發生衝突。

Concurrently, public sentiment indicates a bipartisan shift toward mandatory safety reviews. Data from the AI Policy Institute suggests that a majority of likely voters, including a significant proportion of Republicans, favor government-mandated safety standards over voluntary industry compliance. This societal trend contrasts with the administration's current preference for opt-in reviews, creating a potential misalignment between executive policy and constituent expectations.

與此同時，公眾情緒顯示兩黨均轉向支持強制性安全審查。AI 政策研究所的數據表明，大多數可能投票的選民（包括相當比例的共和黨人）更傾向於政府強制的安全標準，而非業界自願遵守。這一社會趨勢與政府目前偏好的選擇性參與審查形成對比，在行政政策與選民期望之間造成了潛在的不一致。

Internationally, the strategic environment is complicated by the emergence of Chinese capabilities. Reports indicate that Zhipu AI has achieved parity with the Mythos model in vulnerability detection. While the Chinese state maintains a more integrated and restrictive oversight mechanism—evidenced by the historical suppression of Alibaba and the restriction of researcher mobility—the U.S. faces the risk that internal friction between the state and its technological champions may erode its competitive advantage.

在國際上，中國能力的湧現使戰略環境複雜化。報告指出，智譜 AI 在漏洞檢測方面已達到與 Mythos 模型同級的水平。雖然中國政府維持著一個更整合且限制性的監督機制——由歷史上對阿里巴巴的打壓以及對研究人員流動的限制可見——但美國面臨的風險是，州政府與其技術領頭羊之間的內部摩擦可能會削弱其競爭優勢。

The U.S. AI ecosystem remains fragmented, with federal security imperatives clashing with state-level adoption and a growing public demand for formalized regulatory oversight.

美國 AI 生態系統仍然碎片化，聯邦安全指令與州級採納方案發生衝突，且公眾對正式監管監督的需求日益增加。