Goldman Sachs and crypto exchange OKX cut Hong Kong staff off Anthropic’s Claude AI because the model restricts access in Hong Kong and mainland China; Goldman implemented the restriction after a contract review earlier in 2026 while OKX briefly had its corporate account suspended before restoring access. OKX will route Hong Kong employees to other LLMs, spends an estimated $6-8 million a month on large language models, and the incident highlights how provider geography rules complicate AI adoption and workflows across CEXs and financial institutions.

Bearish

Predictions Markets

See what traders are focused on

View analytics →

In Brief

Goldman Sachs and OKX both cut Claude AI access for Hong Kong staff.
OKX's ban followed a brief suspension of its Anthropic account.
Goldman restricted Hong Kong access earlier after a contract review.

Goldman Sachs and OKX have both cut Hong Kong staff off from Anthropic’s Claude AI model. The two restrictions happened months apart, but both trace back to the same geographic access rules.

Anthropic restricts Claude access in Hong Kong and mainland China. OKX and Goldman Sachs each ran into that limit in separate incidents.

OKX’s Ban Followed a Brief Account Suspension

OKX barred staff based in Hong Kong, or traveling through China, from using Claude. The ban followed a brief suspension of OKX’s corporate Anthropic account, Bloomberg reported.

We appreciate Anthropic’s constructive engagement as we work together to address this matter and restore our account access.OKX is a global crypto and technology company, with major offices in California, New York, London, Dubai, Malta, Istanbul, Singapore, and Hong Kong. Given… https://t.co/TeO72UVYZ9
— Star_OKX (@star_okx) August 19, 2026

OKX has since restored the account. The exchange will route Hong Kong employees’ AI requests to other models instead of Claude.

OKX won’t help staff use virtual private networks (VPNs) to reach Claude while traveling through the region.

The timing stands out. OKX has pushed staff to use AI daily and ties that use to performance reviews.

Meanwhile, the exchange reportedly spends $6 million to $8 million a month across multiple large language model (LLM) providers, according to Star Xu, CEO of the exchange. BeInCrypto has reported that OKX and rival exchanges mandate AI tool use for engineering teams.

Goldman Sachs Faced the Same Restriction Months Earlier

Goldman Sachs cut Claude access for its Hong Kong staff earlier in 2026. The bank concluded its Anthropic contract didn’t cover Hong Kong, Bloomberg previously reported.

The restriction also applied to overseas staff visiting Hong Kong. Software engineers made up most of Claude’s user base at the bank.

Goldman’s restriction, in contrast, stemmed from a contract dispute, not a suspension. The bank isn’t retreating from Anthropic overall.

Goldman’s chief information officer, Marco Argenti, has embedded Anthropic engineers within the bank for months. Together, they’re building AI agents for trade accounting and client vetting.

Hong Kong’s government has voiced support for AI adoption among financial firms. However, restrictions from AI developers like Anthropic complicate that push.

Therefore, both firms face the same underlying constraint imposed by Anthropic’s geography rules. The overlap points to a wider pattern.

Washington and Beijing are competing hard for AI supremacy. Chinese AI models keep closing the gap with US developers.

That competition pushes US AI firms to draw harder lines around who can use their technology and where.

Still, neither company has said whether the restrictions reach mainland China-based clients or partners. That question carries more weight with US-China AI talks planned for September.

Read the article at BeInCrypto