How Goldman Sachs and JPMorgan's Claude Ban Reveals the Geopolitics of AI for Hong Kong

JPMorgan Chase followed in June 2026, removing Claude from the internal drop-down menu of approved large language models available to Hong Kong staff . The move was driven by the specific wording of Anthropic's usage terms in its licensing arrangement with the bank, according to three people familiar with the matter .

The root cause is clear in reporting from Reuters and the Financial Times: Anthropic's licensing terms reportedly carve out 'Greater China,' meaning Claude has never been officially supported in Hong Kong . In June 2026, the US government issued an export control directive requiring Anthropic to suspend foreign nationals' access to its newest models, leading to a temporary global disablement . The effect was asymmetric: Hong Kong teams could no longer use a tool their colleagues elsewhere still had.

Why Bernard Chan says rhetoric is not enough

On July 24, 2026, Bernard Chan — a Hong Kong businessman and public figure — published an opinion piece in the South China Morning Post arguing that officials who talk about AI in terms of "the next industrial revolution" or "existential risks" do little to capture how AI actually filters into the ordinary workings of a city like Hong Kong .

His core argument is that abstract rhetoric masks real damage. Claude's removal is not a corporate inconvenience — it is a warning that AI is treated as a strategic asset, not neutral infrastructure . If Hong Kong's teams lose tools that rivals elsewhere retain, the city's competitive position as a global financial centre is materially weakened.

Second, access can disappear overnight. Regional carve-outs and US export controls mean Hong Kong cannot assume continuity of advanced AI tools. Chan argues that city officials must move from grand pronouncements about the 'fourth industrial revolution' to a practical strategy for securing access to the AI technologies that power global finance .

The broader geopolitical picture

These events illustrate that AI governance for Hong Kong is less about regulation and more about supply-chain vulnerability. The city sits at the intersection of US export controls and China's sovereignty claims. Advanced US-origin AI tools are being withdrawn — not because Hong Kong has broken any rules, but because licensing and export-control geography now redraw the map of who gets to use cutting-edge AI. As Chan puts it, the geopolitics of AI are no longer abstract — they are beginning to shape who can use which tools, and on what terms .

Hong Kong's government has acknowledged this risk. A LegCo paper from July 25, 2026, notes that the Innovation and Technology Fund had funded over 390 AI-related projects involving about $1.1 billion in funding as of end-April 2026 . But inside the world's largest banks, the withdrawal of Anthropic's Claude suggests that even well-funded local AI initiatives cannot replace the specific tools that global finance depends on, when those tools are controlled by US licensing and export policies.

Key takeaways

• AI is now a strategic asset, not neutral infrastructure. The bans at Goldman Sachs and JPMorgan show that access to advanced US-origin AI models can be revoked based on geography, licensing terms, and export controls — even for private-sector employees in a major financial hub.
• Hong Kong faces a supply-chain vulnerability in AI. The city cannot assume continuity of access to the tools that power global finance, as Bernard Chan argues. A practical strategy for securing access — rather than abstract rhetoric — is urgently needed.
• The bans targeted Anthropic specifically, not all AI models. ChatGPT and Gemini remained accessible, meaning the restrictions were driven by Anthropic's own licensing carve-outs, not blanket US restrictions on AI in Hong Kong.