JPMorgan Chase removed Anthropic’s Claude models from its approved AI tools list for employees in Hong Kong, the Financial Times reported on June 18. The bank follows Goldman Sachs, which took similar action earlier this month.

The decision reflects JPMorgan’s interpretation of Anthropic’s usage licensing terms under US export control restrictions. Allowing employees in non-US jurisdictions to access Claude models may violate export control compliance requirements tied to the Fable 5 and Mythos 5 restrictions, Reuters reported, citing the FT.

Private Sector Compliance Cascade

This is not direct export control enforcement. JPMorgan was not ordered by the US government to restrict Claude access in Hong Kong. The bank made its own compliance determination: the regulatory ambiguity around Anthropic’s models in foreign jurisdictions creates enough risk that restricting access is the safer corporate posture.

That distinction matters. Government enforcement targets the AI vendor. Private sector compliance cascades target the vendor’s entire customer base in affected geographies. Hong Kong offices of global banks are major users of AI tools for trading analytics, risk assessment, and compliance workflows. Losing access to Claude models in those offices pushes demand toward alternatives: OpenAI’s GPT, Google’s Gemini, and local models.

One Week of Escalation

The JPMorgan decision lands at the end of a week of rapid escalation in the Anthropic export control saga:

  • June 16: Commerce Secretary Howard Lutnick sent a formal letter threatening Anthropic with criminal and civil penalties for unauthorized foreign access to advanced models.
  • June 17: Prediction markets priced a 58-67% chance of Anthropic restoring public Fable 5 access before July 1.
  • June 18: The White House and Anthropic shifted toward negotiating a shared security standards framework. Separately, legal scholars argued the Trump administration’s export control authority may exceed its statutory mandate.
  • June 18: JPMorgan and Goldman Sachs restricted Claude access for Hong Kong staff.

Each step has been covered by NCT. What the JPMorgan move adds: the policy dispute is no longer contained to Washington and Anthropic’s board room. It is changing how the largest financial institutions in the world deploy AI tools across their global operations.

Geographic TAM Contraction

For Anthropic, the downstream effects are immediate. Two of the five largest banks by assets have now restricted Claude in at least one major financial hub. If other multinational enterprises follow the same compliance logic, Anthropic faces reduced demand across Asia-Pacific before any formal export control enforcement has occurred.

The competitive implication is straightforward. OpenAI and Google face no equivalent export control restrictions on their flagship models. Enterprises that need AI tools in Hong Kong, Singapore, and other Asia-Pacific offices now have a compliance reason to choose GPT or Gemini over Claude, independent of model quality.

The open question: if the White House and Anthropic reach a resolution on the security standards framework, how quickly do JPMorgan and Goldman Sachs re-enable Claude access? Corporate compliance changes move slowly. Even a fast resolution in Washington may leave Anthropic locked out of enterprise Hong Kong workflows for months.