根据三名知情人士透露，继 Goldman Sachs 实施类似限制之后，JPMorgan Chase 已禁止其香港员工访问 Anthropic 的 Claude 人工智能模型。这一决定是基于对 Anthropic 授权协议和使用条款的严格解读，该条款将大中华地区排除在外。这项限制对香港作为国际金融中心的地位带来了日益严峻的挑战，特别是在全球竞争对手正迅速采用先进人工智能工具进行软件编码和其他业务的情况下。

尽管香港在历史上的运作并未受到中国大陆「防火长城」（该系统禁用了 OpenAI 的 ChatGPT 和 Claude）的审查，但美国的人工智能开发商仍限制了当地的访问权限以保护其技术。其中一个主要的担忧是「蒸馏」（distillation）技术，即中国本地的从业者可能会利用外国人工智能的输出结果来训练本土模型。虽然一些国际组织先前曾通过在中国境外托管活动并签署全球合同来规避地理限制，但由于这些知识产权和监管问题，访问权限正在收紧。

国家安全考量也起到了关键作用，正如 Anthropic 最近在美国政府指示下暂停了其最先进的 Fable 模型，以限制外籍人士的访问权限所表明的那样。政府官员和金融高管曾特别对 Anthropic 的 Mythos 模型表示担忧，认为其有可能被用来利用全球金融系统中的网络安全漏洞。JPMorgan Chase 和 Goldman Sachs 的举措凸显了在政治敏感地区部署敏感西方技术时，所面临的日益严格的审查和谨慎态度。

JPMorgan Chase has prohibited its Hong Kong staff from accessing Anthropic’s Claude artificial intelligence models, following a similar restriction implemented by Goldman Sachs, according to three individuals familiar with the matter. This decision is based on a strict interpretation of Anthropic’s licensing agreement and usage terms, which exclude the Greater China region. The restriction marks a growing challenge for Hong Kong's position as an international financial hub, particularly as global competitors rapidly adopt advanced AI tools for software coding and other operations.

Although Hong Kong has historically operated without the censorship of mainland China's Great Firewall, where OpenAI's ChatGPT and Claude are banned, US AI developers restrict access to protect their technology. A primary concern is "distillation," a process where local Chinese actors might train native models using foreign AI outputs. While some international organizations previously bypassed geographic restrictions by hosting activities outside China and securing global contracts, access is tightening due to these intellectual property and regulatory concerns.

National security concerns also play a critical role, as demonstrated by Anthropic’s recent suspension of its state-of-the-art Fable model under US government direction to limit access for foreign nationals. Government officials and financial executives had specifically raised alarms regarding the potential of Anthropic’s Mythos model to exploit cybersecurity vulnerabilities within the global financial system. The actions by JPMorgan Chase and Goldman Sachs highlight the increasing scrutiny and caution surrounding the deployment of sensitive Western technologies in politically sensitive regions.