The New Digital Iron Curtain Is a Terms of Service Agreement

JPMorgan just blocked a top AI model for its Hong Kong staff. It's not a bug. It's a sign that the global internet is breaking apart, one corporate policy at a time.

Wall Street is quietly building a new kind of firewall in Hong Kong. JPMorgan Chase just cut its bankers there off from Anthropic's Claude, one of the most capable large language models on the market. This isn't an outage or a technical glitch. It's a deliberate choice. The decision follows a similar move by Goldman Sachs just months prior, as reported by outlets like Reuters. When two of the world's biggest banks independently decide a leading AI tool is too risky for one specific city, it’s not about the tool. It's about the city, and the invisible lines of jurisdiction that are being drawn through the global network.

The block has nothing to do with Claude’s performance and everything to do with its license. Anthropic is an American AI lab backed by billions from Amazon and Google, and its operations are subject to intensifying US government scrutiny. The core of the problem is data sovereignty. A query typed by a banker in a Hong Kong office gets processed on servers that could be anywhere, introducing legal ambiguity. The banks' risk and compliance departments, staring at the shifting legal ground in Hong Kong and strict US export policies, are making a judgment call. They’ve read the fine print and concluded that the potential for data to fall under the purview of mainland Chinese authorities creates an unacceptable liability. This is a failure mode of legal code, not software code.

The money trail tells a story of strategic retreat. For JPMorgan and Goldman Sachs, the immediate loss is a productivity hit for their Hong Kong divisions, which now can't use the same advanced tools as their counterparts in London or New York. For Anthropic, it's a lost beachhead in Asia's premier financial hub. The real winner is geopolitical fragmentation itself. As the Financial Times first reported the JPMorgan move, it underscored a larger trend: the balkanization of the enterprise tech stack. The chief beneficiaries may ultimately be state-aligned Chinese AI developers, who face a suddenly less competitive market. What’s truly at stake is Hong Kong’s viability as a neutral clearinghouse where capital, data, and talent can mix freely between East and West. That identity is being dismantled, one SaaS contract at a time.

This is a preview of the next five years. Expect to see more of these compliance-driven blockades, creating fractured digital workplaces inside multinational corporations. We are rapidly heading toward a world with two parallel enterprise stacks — one compliant with US rules, one for China's sphere of influence — and global firms will have to pay to maintain both. The real test will come not from another policy update, but from the first major international incident traced back to a US tech company's data handling in the region. The question is no longer whether your company will use AI. It's whether your company will be allowed to use the same AI in every city where it does business.