美国禁止联邦机构使用 DeepSeek，引发中方强烈反对

发生了什么？ DeepSeek 可能是目前最受欢迎的非美国 AI 公司之一，但它正被美国联邦民事机构拒之门外。华盛顿已下令从政府 IT 系统中移除这家中国公司的 AI 产品，并阻止未来的采购，相关限制还包括使用其技术构建的第三方工具。不出所料，北京方面对此表示强烈不满。

美国国土安全部（DHS）的通知于 10 月 19 日公开发布，早于今日在《联邦公报》（Federal Register）上的正式刊载。据美联社（AP）报道，该指令的基础命令实际上可追溯至 7 月底，这意味着这只是对早前决定的一项公开披露，而非本周签署的新禁令。

该指令涵盖 DeepSeek、其关联机构以及使用 DeepSeek 实现方案开发的产品和服务。最后这一项规定意味着，该指令不仅限于阻止对该公司聊天机器人的直接访问，联邦机构还必须全面审查所采购软件的底层技术支持。

国土安全部援引了《2018年联邦采购供应链安全法》（Federal Acquisition Supply Chain Security Act of 2018）。其陈述的担忧包括 DeepSeek 的服务器位于中国，中国当局可能借此访问敏感的美国信息。该命令还补充称，DeepSeek 的 AI 模型“构成了多重威胁载体”，并警告潜在的恶意软件及其他安全缺陷，指出有多项公开研究表明 DeepSeek 的 iOS 应用程序中存在多个严重漏洞。

中国外交部发言人毛宁在周二作出反击，指责华盛顿泛化国家安全概念，将科技和经贸问题政治化。毛宁补充指出，中国将“坚决维护中国企业的合法正当权益”。DeepSeek 未立即回应美联社的置评请求。

各行政机构自命令发布之日起有 60 天时间向国土安全部提交书面正当理由申请豁免，包括争取更长的合规缓冲期。

对这家中国公司来说，类似处境并不陌生。2025 年 1 月，美国海军曾警告官兵不要使用 DeepSeek。然而，海军随后澄清，该通告仅反映了针对将公共生成式 AI 用于官方工作的既有限制（该限制同样适用于 ChatGPT），并非专门针对某一家中国聊天机器人出台的新政策。

美国本土的 AI 企业同样遭遇过限制。2024 年 4 月，出于对国会数据流向未经批准的云服务的担忧，美国众议院封禁了 Copilot。但该立场在 2025 年 9 月发生改变，众议院宣布在具备额外法律和数据保护措施的前提下部署 Copilot。此外，还有围绕 Anthropic 的一系列风波。

尽管 DeepSeek 目前并未面临全美范围内的禁令，但该指令确实会直接影响那些向美国联邦民事机构销售由其模型驱动的 AI 软件的企业。

What just happened? DeepSeek might be one of the most popular non-US AI companies right now, but it's being shown the door across US federal civilian agencies. Washington has ordered the removal of the Chinese firm's AI products from government IT systems and blocked future purchases, and the restrictions include third-party tools built using its technology. Beijing, unsurprisingly, isn't happy.

The Department of Homeland Security's notice was made public on October 19, ahead of today's Federal Register publication. The underlying order dates back to late July, according to AP, making this a public disclosure of an earlier decision rather than a ban signed this week.

The directive covers DeepSeek, its affiliates, and products and services developed using DeepSeek implementations. That last part means the directive is more than just blocking access to the company's chatbot. Agencies also have to consider what powers the software they buy.

DHS invoked the Federal Acquisition Supply Chain Security Act of 2018. Its stated concerns include DeepSeek's servers being located in China and the possibility that Chinese authorities could access sensitive US information. The order added that DeepSeek's AI models "present multiple threat vectors." It also warned of potential malware and other security flaws, pointing to publicly available research that found several critical vulnerabilities in DeepSeek's iOS app.

Chinese Foreign Ministry spokesperson Mao Ning hit back on Tuesday, accusing Washington of stretching national security arguments and politicizing technology and trade. Ning added that China will "will resolutely safeguard the legitimate rights and interests of Chinese companies." DeepSeek did not immediately respond to AP's request for comment.

Executive agencies have 60 days from the order's issuance to submit a written justification to DHS seeking an exception, including more time to comply.

This is familiar ground for the Chinese firm. In January 2025, the US Navy warned personnel against using DeepSeek. However, the military branch subsequently clarified that the message reflected existing restrictions on publicly accessible generative AI for official work, which also covered ChatGPT. It wasn't a newly invented policy aimed solely at one Chinese chatbot.

American AI companies have faced restrictions, too. The House of Representatives banned Copilot in April 2024 over concerns about congressional data reaching unapproved cloud services. That position changed in September 2025, when the House announced a Copilot rollout with additional legal and data protections. There's also the whole Anthropic saga.

While DeepSeek isn't facing a nationwide ban, the directive does affect companies selling AI software powered by its models to US federal civilian agencies.