JPMorgan has restricted access to Anthropic’s Claude models for employees in Hong Kong. The Financial Times, citing three people familiar with the matter, reported that staff in the city can no longer choose Claude from the bank’s internal list of approved large language models.
According to the report, the change is tied to wording in Anthropic’s licensing agreement. One person familiar with the decision said JPMorgan relied on terms that govern where the models may be used. Those terms reportedly exclude Greater China, including Hong Kong. JPMorgan declined to comment. Anthropic has not issued a formal statement, though it previously told the Financial Times that Claude had never been officially supported in Hong Kong.
Goldman Sachs had already made a similar move
The restriction follows a similar step by Goldman Sachs earlier this year. The Financial Times had previously reported that Goldman blocked bankers in Hong Kong from using Anthropic models after concluding that the company’s terms of service excluded usage across Greater China.
That matters because Hong Kong has long operated with fewer internet restrictions than mainland China. OpenAI’s ChatGPT and Anthropic’s Claude are unavailable in mainland China due to a mix of company policy and China’s internet controls. In Hong Kong, international firms have often obtained access to frontier AI systems through global enterprise agreements and infrastructure hosted outside China.
Access limits raise pressure on Hong Kong’s AI tool stack
The Financial Times said the new limits at major financial institutions have renewed concerns about whether Hong Kong can stay competitive as AI tools become more embedded in software development, research, and financial-services workflows. The issue is no longer just about one model. It touches how firms source and deploy advanced systems across regions.
U.S. AI companies are also facing closer scrutiny over how advanced models are used outside the United States. Industry observers and policymakers have warned that foreign users could rely on frontier systems to speed up domestic AI development through model distillation, pushing companies to tighten geographic controls.
Anthropic is also dealing with export controls and litigation
The banking restrictions came less than a week after Anthropic suspended access to its newly released Fable 5 and Mythos 5 models. On June 13, the company said it had disabled both systems after receiving a U.S. government export-control directive. Anthropic said authorities instructed it to block access for all foreign nationals, including foreign-national employees working inside the United States.
At the time, Anthropic said officials were concerned about a possible jailbreak technique that could enable the models to identify or repair software vulnerabilities. The company disputed the significance of that issue and said it believed the government action may have resulted from a misunderstanding.
Just two days later, Anthropic was named in a proposed class-action lawsuit in the U.S. District Court for the Northern District of California. The complaint alleges that subscribers to the company’s $100-per-month Max 5x and $200-per-month Max 20x Claude plans received substantially less usage than the marketing materials suggested. Plaintiff Karl Kahn is seeking class-action status on behalf of customers who have paid for premium Claude subscriptions since April 2024, arguing that the usage limits did not match the multipliers promoted for those plans.
Those disputes landed shortly after Anthropic publicly called for tighter oversight of frontier AI systems. In its June 11 proposal, “Policy on the AI Exponential,” the company urged governments to set testing requirements, independent evaluations, cybersecurity standards, and enforcement mechanisms for the most advanced AI models. Anthropic argued that frontier systems can introduce biological, cybersecurity, and operational risks.