Key Takeaways
- Anthropic’s Claude AI has been disabled for Goldman Sachs employees based in Hong Kong following a recent contract assessment
- Goldman’s internal AI platform continues to support alternative tools such as ChatGPT and Gemini
- A strict interpretation of Goldman’s agreement with Anthropic led to the determination that Hong Kong staff should avoid all Anthropic offerings
- Anthropic’s website does not designate Hong Kong as an officially supported region
- Financial institutions and regulators worldwide are examining Anthropic’s latest Mythos model amid concerns about potential banking sector vulnerabilities
Investment banking powerhouse Goldman Sachs has discontinued access to Anthropic’s Claude artificial intelligence platform for its workforce stationed in Hong Kong, a person with firsthand knowledge of the matter has revealed.
The individual informed Reuters that Hong Kong-based personnel previously utilized Claude via a proprietary AI infrastructure. Within the past several weeks, this capability has been withdrawn.
Competing AI solutions, including Google’s Gemini platform and OpenAI’s ChatGPT system, remain accessible to employees through the identical internal network.
The Financial Times initially disclosed this development, referencing sources with knowledge of the situation. Goldman Sachs refused to provide commentary. Anthropic has not yet issued a response to inquiries.
As reported by the FT, Goldman arrived at this determination following a rigorous examination of its contractual obligations with Anthropic. The financial institution engaged in discussions with Anthropic prior to establishing that personnel in Hong Kong should discontinue use of any Anthropic solutions.
An Anthropic representative informed the FT that Claude has never received official “support” status in Hong Kong. The organization’s online presence does not include Hong Kong among territories where its API or Claude.ai platform are formally accessible.
The Hong Kong Situation Explained
Hong Kong occupies a uniquely complex position regarding American AI technologies. While AI systems developed by U.S. corporations face prohibition in mainland China, Hong Kong has generally maintained openness, with availability determined by individual corporate policies.
This development unfolds against a backdrop of escalating U.S.-China friction surrounding artificial intelligence capabilities and information security. An anticipated summit between Presidents Donald Trump and Xi Jinping scheduled for mid-May is projected to address these matters.
The U.S. administration released a worldwide alert last week regarding suspected AI intellectual property theft by Chinese AI entities. During 2024, OpenAI similarly implemented measures to prevent China-based traffic to its API platform due to apprehensions that its technologies might be exploited by Chinese rivals for model training purposes.
Reuters was unable to verify whether additional financial institutions or enterprises have imposed comparable restrictions on Claude availability in Hong Kong.
Mythos Model Triggers Regulatory Scrutiny
Goldman’s decision coincides with heightened interest from international banking institutions and financial oversight bodies regarding Anthropic’s most recent AI system, Mythos.
The Hong Kong Monetary Authority confirmed to Reuters that it has reached out to numerous prominent banks to gather intelligence on recent Mythos-related developments. The regulator additionally reminded these institutions to refresh their risk evaluations and implement suitable protective measures.
Reuters documented that certain lending organizations were already evaluating additional security protocols in reaction to cybersecurity challenges associated with sophisticated AI technologies.
During February, Goldman Sachs Chief Information Officer Marco Argenti indicated the institution was collaborating with Anthropic to create AI-driven agents designed to automate an expanding array of operational tasks.
The Hong Kong government has not provided a response to Reuters’ request for statement. The HKMA declined to address Goldman’s particular action.
