2025年9月5日，美國人工智慧公司Anthropic宣布一項重磅政策，立即停止向中資控股超過50%的實體提供 Claude 服務。這一舉措不僅針對中國大陸的公司，還包括設立於海外的子公司，甚至涵蓋那些間接使用Claude的中資背景機構。消息一出，立刻在全球科技與資本市場引發廣泛關注。

此次政策的範圍相當嚴格。凡是由中國資本直接或間接控股超過50%的企業，不論註冊地在中國、東南亞，甚至美國本土，只要符合條件，皆被列入限制對象。與此同時，Anthropic也將俄羅斯、伊朗、北韓等美國長期視為「對手國家」的實體納入封禁名單，體現出這項政策背後強烈的地緣政治色彩。

至於封禁理由，Anthropic在聲明中強調，此舉是出於對 法律、監管與國家安全風險 的考量。公司擔憂中資企業可能藉由「模型蒸餾」等技術，將Claude的能力轉化為自身資產，並進一步應用於軍事、情報或其他可能威脅美國國家利益的領域。此外，Anthropic也隱晦指出，中資科技巨頭或許會利用這些技術在全球市場上與美國企業展開更直接的競爭，這同樣是風險來源之一。

在影響層面上，此政策對中國科技產業的衝擊不容小覷。像字節跳動、騰訊、阿里巴巴等擁有跨國業務的中資巨頭，若其部分專案或應用仰賴Claude API，將面臨服務被迫中斷的風險。這不僅可能導致海外研發或合作專案受挫，還可能使相關業務需緊急轉向替代解決方案。更重要的是，Anthropic明確封堵透過雲端服務或第三方平台「曲線使用」Claude的途徑，讓中資企業難以藉由代理商或合作夥伴繞過封禁。

這一事件不僅是一家AI公司的商業決策，更深層地反映了當前中美科技博弈的激烈態勢。AI作為戰略性技術，其發展與控制權已經上升至國家安全高度。Anthropic的封禁舉措，很可能成為其他美國科技公司效仿的先例，也將迫使中資企業加快培養自主研發能力，尋找替代方案，進一步加劇全球人工智慧領域的分化與對立。

On September 5, 2025, the American artificial intelligence company Anthropic announced a major policy shift, immediately halting the provision of Claude services to entities with more than 50% Chinese ownership. This decision targets not only companies based in mainland China, but also overseas subsidiaries and institutions with Chinese capital backgrounds that indirectly use Claude. The news has quickly drawn significant attention in the global tech and capital markets.

The scope of this policy is extremely strict. Any enterprise directly or indirectly controlled by Chinese capital above the 50% threshold—whether registered in China, Southeast Asia, or even within the United States—will be subject to the restrictions. At the same time, Anthropic has also included entities in countries long regarded by the U.S. as “adversaries,” such as Russia, Iran, and North Korea, further highlighting the geopolitical undertones of this move.

As for the reasons behind the ban, Anthropic emphasized that it stemmed from concerns over legal, regulatory, and national security risks. The company fears that Chinese enterprises might use techniques such as “model distillation” to extract Claude’s capabilities, repurposing them for military, intelligence, or other fields that could threaten U.S. national interests. Anthropic also hinted at worries that Chinese tech giants could leverage these technologies to compete more directly with American firms in the global marketplace, which itself poses a strategic risk.

The impact on China’s tech sector is significant. Major companies such as ByteDance, Tencent, and Alibaba—all of which run multinational projects—could see operations disrupted if certain services or applications rely on Claude’s API. This may not only stall overseas R&D or collaborative ventures, but also force businesses to urgently shift toward alternative solutions. More importantly, Anthropic explicitly blocked potential “workarounds” via cloud services or third-party platforms, preventing Chinese firms from indirectly accessing Claude through intermediaries.

This event represents far more than a single AI company’s business decision—it underscores the intensifying technology rivalry between China and the U.S. As AI increasingly becomes a strategic technology, its development and control have escalated into matters of national security. Anthropic’s ban may set a precedent for other American tech firms to follow, while simultaneously pressuring Chinese enterprises to accelerate self-reliance in AI R&D and seek alternatives. In the long run, this move may deepen fragmentation and confrontation within the global AI landscape.