（版權所有：《創新律師》）

6月，美國總統唐納德·特朗普簽署第14409號行政令《促進先進人工智能創新與安全》。這項具有里程碑意義的政策，重新定義了美國人工智能發展和網絡安全國家戰略。7月中旬，白宮正式啟動“金鷹倡議”，這是根據行政令第2（d）條授權建立的人工智能驅動網絡安全信息共享平台，它標誌着美國開始以新方式主導全球人工智能格局。

第14409號行政令及“金鷹”行動機構遠非單純的國內監管調整，它們加劇了本已激烈的全球人工智能競爭，並促使世界各國加大定向投資、戰略產業規劃以及安全化人工智能部署的投入力度。曾經的技術優勢之爭已演變為涉及地緣政治、經濟和軍事的多維度較量，若國際社會未建立基於規則的合作框架來平衡創新與安全，這種對抗有可能把全球科技生態系統分割為相互對立的陣營。

謀求主導地位

第14409號行政令是特朗普政府歷時一年、圍繞“美國優先”原則重新定位美國人工智能政策的成果，旨在扭轉所謂拜登政府對科技行業的過度監管限制。其目的是：通過與私營部門合作而非施加繁重的自上而下監管，保持美國在人工智能創新方面的全球領先地位；通過強化關鍵基礎設施以抵禦人工智能驅動的網絡威脅，在不受約束的技術發展與國家安全之間達成審慎平衡；保護美國的知識產權和技術優勢，防範其他國家竊取、複製美國人工智能發展成果的敵對行為。

第14409號行政令和“金鷹倡議”是特朗普政府精心謀劃的戰略舉措，旨在利用美國無與倫比的私營部門人工智能優勢——以OpenAI、Anthropic和谷歌等公司為代表——同時藉助政府協調的網絡安全合作，來保護美國關鍵基礎設施免受外國對手侵害。通過將人工智能創新直接與國家安全掛鉤，這項政策不僅維護美國技術優勢，更將人工智能網絡安全武器化，使國內政策工具成為在全球彰顯地緣政治影響力的手段。

全球人工智能競爭

美國單方面升級以人工智能為核心的網絡安全和技術競爭，已引發全球性連鎖反應，主要經濟體和中等強國紛紛推出協調一致、資金充足的戰略，以提升自身的人工智能能力。由此出現了一個分叉的全球人工智能競賽：一邊是美國及其盟友，它們推行的模式是將私營部門主導的創新與政府保障的技術優勢相結合；另一邊是中國及其他國家，它們正在推進以國家為中心的全鏈條人工智能戰略，以減少對西方技術的依賴，並挑戰美國霸權。

對美國而言，“金鷹”遠不止是一項網絡安全措施，更是一種鞏固全球人工智能主導地位的戰略工具。通過實現聯邦政府與美國頂尖人工智能開發企業之間漏洞情報的集中共享，該舉措旨在確保美國公司在全球部署其模型的同時，仍能保持決定性的防禦技術優勢。第14409號行政令拒絕給人工智能開發企業施加新監管，進一步強化了這一優勢：它消除了可能拖慢國內創新的合規成本，而同時，政府繼續對高性能人工智能芯片和前沿模型技術實施嚴格的出口管制，以阻止敵對國家獲取美國的先進成果。其總體目標很明確，那就是確保美國在人工智能創新以及支撐其全球部署的網絡安全框架方面，繼續保持無可匹敵的領先地位。

中國的對策

中國奉行“雙輪驅動”戰略，即依託本土半導體製造業，打造自主的人工智能硬件能力，同時利用人工智能軟件進步，來升級傳統產業。從製造業到醫療保健等領域都在進行這樣的實踐，這是中國應對美國升級行動的一部分。這是一項由國家主導、舉全國之力推進的人工智能戰略，目的是通過自主創新，降低對西方的技術依賴。

今年7月，中國國家發改委正式發佈《人工智能合作發展行動計劃》，這是一項綜合性政策，涵蓋了八個戰略重點：

• 擴大優質人工智能數據供給；

• 促進智能算力普惠可及；

• 構建開源生態共享體系；

• 推動人工智能產業深度賦能；

• 培養專業技術人才；

• 共建全球技術規則與標準；

• 協作開展人工智能安全治理；

• 推進人工智能“向善”公共實踐。

俄羅斯及其他參與方

俄羅斯的人工智能戰略高度聚焦於核心國家安全需求以及西方制裁的限制，重點關注進口替代和軍事領域人工智能應用。今年7月，俄羅斯政府確認了一項每年717億盧布的國內人工智能產業新支持方案。該計劃包括對人工智能模型開發者的稅收優惠，以及為新數據中心提供電網優先接入權，這對在能源受限的俄羅斯科技生態系統中運營的企業來說是一個關鍵優勢。此外，通過創新援助基金會的“Start-II”聯邦競賽，小型科技企業可以獲得最高3000萬盧布的資助，用於開發面向國內市場的人工智能解決方案，這些項目的關鍵要求是必須替代因國際制裁而無法再獲取的進口軟件或硬件。

韓國已將自己定位為全球人工智能競賽中的關鍵搖擺力量，利用其世界領先的半導體製造能力，開闢出一席之地，成為重要的人工智能基礎設施供應商。今年3月，韓國總統直屬的國家人工智能戰略委員會宣布2026年為“人工智能G3飛躍之年”，並公布了總計9.9萬億韓元（約合71億美元）的人工智能預算，幾乎是上一年撥款的三倍，占該國國家研發總預算的28.6%。

日本則專註於“主權人工智能”能力和實體人工智能集成。具體而言，是利用人工智能驅動的機械人，來解決長期存在的勞動力短缺問題，並振興傳統製造業。今年，日本政府公布了一項1萬億日元的主權人工智能計劃，這是其更廣泛的370萬億日元（2.3萬億美元）國家增長戰略的一部分。該計劃將推動擴大國內半導體製造，資助理化學研究所和國家信息學研究所的人工智能研究中心（這些中心正在開發供政府和企業使用的本土大語言模型），支持中小企業使用人工智能工具與系統。

美國發起的“金鷹倡議”凸顯了全球人工智能競爭中的一條決定性分界線：美國以私營部門為主導、聚焦安全的模式，與中國國家統籌、全鏈條創新的戰略形成直接對立。包括韓國和日本在內的中等強國正利用自身核心工業優勢——韓國的半導體、日本的精密製造——在全球人工智能供應鏈中開闢獨立的細分市場，以應對這一分化格局。與此同時，俄羅斯等國則奉行有針對性、抗制裁的人工智能戰略，優先考慮國家安全而非全球市場份額。

這種分叉式競爭正在分裂全球技術供應鏈，加速人工智能的安全化進程，並將世界推向一種新型技術冷戰。這種冷戰有可能使人工智能成為永久性全球衝突的根源，而非促進共同進步的共享工具。

正確的道路

由美國“金鷹倡議”及其他國家戰略引發的全球人工智能競賽正處於一個臨界點。如果沒有協調一致的國際行動，零和博弈不僅會將全球人工智能生態系統分裂為相互競爭、互不兼容的陣營，還會加劇網絡衝突，擴大全球數字鴻溝，並使人工智能驅動的關鍵基礎設施面臨更大的風險。為確保人工智能兌現全球共同繁榮的承諾，同時減輕人工智能引發衝突或被濫用的風險，國際社會應做到以下三點：

第一，用建設性、有針對性的國際合作取代惡性競爭。人工智能驅動的網絡威脅——從針對關鍵基礎設施的漏洞利用，到由前沿模型驅動的虛假信息傳播——本質上是無國界的。美國人工智能模型中的一個漏洞，可能被亞洲或歐洲的行為體利用。對中國人工智能數據中心的網絡攻擊，可能擾亂支撐各大洲經濟的全球供應鏈。這一現實意味着，沒有任何一個國家，即使是像美國這樣技術先進的國家，能夠僅憑單邊行動來保障其人工智能系統或關鍵基礎設施的安全。

第二，推動開源人工智能，以對抗技術霸權，縮小全球鴻溝。當前人工智能競爭最危險的副作用之一，是人工智能技術力量日益集中在少數西方科技巨頭和富裕發達國家手中。封閉的、專有的人工智能模型——由OpenAI和Anthropic等公司開發和掌控——賦予美國前所未有的影響力，可以左右人工智能技術在全球範圍內的使用方式，而許多發展中國家卻沒有構建自身前沿人工智能系統的技術或財力。這種動態正在催生新的數字殖民主義，迫使發展中國家從醫療保健到國家安全等各個領域，都依賴美國控制的人工智能工具。這擴大了全球數字鴻溝，加劇了現有的全球不平等。

第三，建立以聯合國為中心的多邊治理體系，以強制執行全球人工智能規範。為使國際合作和開源人工智能協作具有可持續性，國際社會必須以普遍適用、具有法律約束力的治理框架，來取代當前分散且以國家層面為主的人工智能監管體制。

目前，全球人工智能規則是由相互衝突的各國法律拼湊而成。例如，歐盟《人工智能法案》對高風險人工智能系統施加了嚴格的合規要求，美國則依賴自願的公私合作，而中國的人工智能法規是根據國家主導的發展模式量身定製的。這種碎片化給全球科技公司帶來高昂合規負擔，讓不良行為者有可鑽的空子，也絲毫未能約束正在推動當前人工智能競賽的民族主義技術政策。

聯合國是唯一具備合法地位、有能力引導建立協調一致的多邊人工智能治理框架的全球性機構。今年7月，聯合國召開首次全球人工智能治理對話，彙集了所有193個成員國的代表，討論共同的人工智能風險和機遇。這是關鍵的第一步。但對話只是一個討論平台，而非規則制定機構，缺乏強制執行具有約束力的全球規範的權力。

為推動這一進程，聯合國大會應將現有聯合國人工智能高級別諮詢機構正式升級為常設專門監管機構，負責起草一項具有全球約束力的人工智能安全公約，涵蓋人工智能系統從開發、測試到部署和持續使用的整個生命周期。

人工智能革命太過重要，風險也過於巨大，不能放任其在不受監管的國家競爭中發展。通過以合作取代對抗，將開源人工智能作為全球公共產品加以推進，並在聯合國主持下建立一個普遍適用、有章可循的治理體系，世界才能確保人工智能成為促進全球繁榮的共同工具，而不是下一場國際衝突的主戰場。

選擇很簡單：要麼世界共同治理人工智能，要麼人工智能驅動的競爭進一步分裂世界。每個國家都應認真思考這個問題。

(Copyright: The Innovation Attorney)

In June, U.S. President Donald Trump issued Executive Order 14409, “Promoting Advanced Artificial Intelligence Innovation and Security”—a landmark policy that redefines America's national strategy for AI development and cybersecurity. In mid-July, the White House formally launched the Gold Eagle Initiative—an AI-powered cybersecurity clearinghouse mandated under Section 2(d) of the earlier order and marking the start of a new American approach to dominating the global AI landscape.

Far from being a mere domestic regulatory adjustment, Executive Order 14409 and its Gold Eagle operational arm have escalated what was already fierce global AI competition and have driven nations around the world to double down on targeted investments, strategic industrial planning and securitized AI deployment. What was once a contest over technological superiority has evolved into a multifaceted geopolitical, economic and military rivalry, one that threatens to split the global tech ecosystem into competing blocs unless the international community adopts a cooperative, rules-based framework to balance innovation with security.​

Seeking to dominate

EO 14409 is the culmination of the Trump administration's yearlong effort to reframe American AI policy around his “America first” principle, reversing what it viewed as the Biden administration's overreaching regulatory constraints on the tech sector. Its purposes are to preserve America's global lead in AI innovation by partnering with the private sector rather than by imposing burdensome top-down regulation; to strike a deliberate balance between unfettered tech development and national security by hardening critical infrastructure against AI-driven cyberthreats; and to protect American intellectual property and technological superiority from adversarial efforts by other countries to steal or replicate U.S. AI advancements.​

EO 14409 and the Gold Eagle Initiative are part of a deliberate strategic calculation by the Trump administration to leverage America's unrivaled private-sector AI dominance—led by companies such as OpenAI, Anthropic and Google—while using government-coordinated cybersecurity collaboration to insulate America's critical infrastructure from foreign adversaries. By tying AI innovation directly to national security, the policy does not merely defend U.S. technological superiority; it weaponizes AI cybersecurity, turning a domestic policy tool into a means of asserting global geopolitical leverage.​

Global AI competition

America's unilateral escalation of AI-centric cybersecurity and technological competition has triggered a cascading global response, with major economies and middle powers alike rolling out coordinated, well-funded strategies to advance their own AI capabilities. What has emerged is a bifurcated global AI race: On one side, the United States and its treaty allies are pursuing a model that pairs private-sector-led innovation with government-secured tech dominance; on the other, China and other nations are advancing state-centered, full-chain AI strategies to reduce reliance on Western technology and challenge U.S. hegemony.​​

For the U.S., Gold Eagle is far more than a cybersecurity measure; it is a strategic instrument to consolidate global AI dominance. By centralizing vulnerability intelligence-sharing between the federal government and America's leading AI developers, the initiative aims to ensure that U.S. companies maintain a decisive defensive technological edge even as they deploy their models globally. EO 14409's refusal to impose new regulations on AI developers further fuels this advantage: It eliminates compliance costs that could slow domestic innovation, while the administration continues to enforce strict export controls on high-performance AI chips and frontier model technology to block adversarial nations from accessing U.S. advancements​. The overarching goal is clear: to ensure that the United States remains the unrivaled leader in both AI innovation and the cybersecurity frameworks that enable global deployment.​

China's approach​

China has followed a strategy of “double wheel driving”—that is, leveraging homegrown semiconductor manufacturing to build independent AI hardware capacity, while using AI software advancements to upgrade traditional industries. This is happening in sectors from manufacturing to healthcare as part of China's response to the U.S. escalation. It's a state-led, whole-of-nation AI strategy focused on indigenous innovation to reduce technological dependence on the West.

In July, China's National Development and Reform Commission formally issued its AI Cooperation Development Action Plan, a comprehensive policy that lays out eight strategic priorities:

• Expanding high-quality AI data supplies;

• Lowering barriers to intelligent computing power access;

• Nurturing a shared open-source ecosystem,

• Accelerating AI industrial integration

• Cultivating specialized tech talent;

• Aligning global tech rules and standards;

• Collaborating on AI security governance; and

• Advancing AI's “for good” public applications​.

Russia and others in the mix​

Russia's AI strategy is narrowly tailored to its core national security needs and the constraints of Western sanctions, with a focus on import substitution and military AI applications. In July, the Russian government confirmed a new support package for the domestic AI industry worth 71.7 billion rubles annually. The plan includes tax incentives for AI model developers and priority access to electrical grids for new data centers—a critical advantage for firms operating in Russia's energy-constrained tech ecosystem​. Through the Innovation Assistance Foundation's “Start-II” federal competition, small tech enterprises can access grants of up to 30 million rubles to develop AI solutions for domestic use, with the key requirement that these projects replace imported software or hardware that Russian entities can no longer access due to international sanctions​.​

South Korea has positioned itself as a critical swing state in the global AI race, leveraging its world-leading semiconductor manufacturing capability to carve out a niche as a key AI infrastructure supplier. In March, the country's Presidential National AI Strategy Committee declared 2026 the “AI G3 Leap Year” and unveiled a total AI budget of 9.9 trillion won ($7.1 billion)—nearly triple the previous year's allocation, That represents 28.6 percent of the country's total national research and development budget​.​

Japan is focusing on “sovereign AI” capacity and physical AI integration—specifically, using AI-powered robots to address its longstanding labor shortage and revitalize the traditional manufacturing sector. This year, the Japanese government unveiled a 1 trillion yen sovereign AI plan as part of its broader 2.3 trillion national growth strategy. The package allocates 800 billion yen to expand domestic semiconductor manufacturing. Another 400 billion yen funds AI research centers at RIKEN and the National Institute of Informatics, which are developing homegrown large language models for government and corporate use. And 500 billion yen is earmarked to support small and medium-sized enterprises in adopting AI tools and systems​.​

The Gold Eagle Initiative undertaken by the United States has sharpened the defining fault line in global AI competition: The private-sector-led, security-focused model in the U.S. stands in direct opposition to China's state-coordinated, full-chain innovation strategy. Middle powers, including South Korea and Japan are navigating this divide by leveraging their core industrial strengths—semiconductors for South Korea, precision manufacturing for Japan—to secure independent niches in the global AI supply chain. Meanwhile, countries such as Russia are pursuing targeted, sanction-resistant AI strategies that prioritize national security over global market share.

The bifurcated competition is fragmenting global tech supply chains, accelerating the securitization of AI and pushing the world closer to a new kind of technological cold war—one that threatens to make AI a source of permanent global conflict, rather than a shared tool for collective progress.​

The right path

The global AI race triggered by America's Gold Eagle Initiative and other national strategies is at a tipping point. Without coordinated international action, zero-sum rivalry will not only divide the global AI ecosystem into competing, incompatible blocs but will also escalate cyber conflicts, widen the global digital divide and put AI-powered critical infrastructure at ever-greater risk. To ensure AI delivers on its promise of global shared prosperity while mitigating the risks of AI-driven conflict or misuse, the international community should do the following three things:

First, replace malignant rivalries with constructive, targeted international cooperation​. AI-driven cyberthreats—from vulnerability exploits against critical infrastructure to disinformation campaigns powered by frontier models—are inherently borderless. A vulnerability in an American AI model can be exploited by actors in Asia or Europe; a cyberattack on a Chinese AI data center can disrupt global supply chains that support economies on every continent. This reality means that no single nation, even one as technologically advanced as the United States, can secure its AI systems or critical infrastructure through unilateral action alone.​

Second, promote open-source AI to counter technological hegemony and narrow the global divide​. One of the most dangerous side-effects of the current AI competition is the growing concentration of AI technological power in the hands of a small number of Western tech giants and wealthy developed nations. Closed, proprietary AI models—developed and controlled by companies like OpenAI and Anthropic—give the United States unprecedented influence over how AI technology is used globally, while many developing nations lack the technical or financial resources to build their own frontier AI systems. This dynamic is creating a new digital colonialism in which developing countries are forced to rely on U.S.-controlled AI tools for everything from healthcare to national security. This widens the global digital divide and exacerbates existing global inequalities.​

Third, establish a UN-centered multilateral governance system to enforce global AI norms. For international cooperation and open-source AI collaboration to be sustainable, the global community must replace the current fragmented, national-level AI regulatory regimes with a universal, legally binding governance framework.

Currently, global AI rules are a patchwork of conflicting national laws. For example, the European Union AI Act imposes strict compliance requirements on high-risk AI systems; the United States relies on voluntary public-private collaboration; and China's AI regulations are tailored to its own state-led development model. This fragmentation creates costly compliance burdens for global tech companies, creates loopholes that bad actors can exploit and does nothing to constrain the nationalistic tech policies that are driving the current AI race.​

The United Nations is the only legitimate global body capable of shepherding a coordinated, multilateral AI governance framework. In July, the UN convened its first Global Dialogue on AI Governance, bringing together representatives from all 193 member states to discuss shared AI risks and opportunities. This was a critical first step—but the dialogue is a forum for discussion, not a rule-setting body; it lacks the authority to enforce binding global norms.

To move forward, the UN General Assembly should formalize the existing UN AI Governance Advisory Board into a permanent, specialized regulatory body tasked with drafting a globally binding AI security convention that covers the entire lifecycle of AI systems, from development and testing to deployment and ongoing use.​

The AI revolution is too important, and its risks too great, to be left to unregulated national competition. By replacing rivalry with cooperation, advancing open-source AI as a global public good, and establishing a universal, rules-based governance system under the UN's auspices, the world can ensure that AI becomes a shared tool for global prosperity, not the next great battlefield of international conflict.

The choice is simple: either the world governs AI together, or AI-driven competition further divides the world. Every country should consider this issue seriously.