A major international bank has moved to block staff from using the Claude artificial intelligence chatbot in an Asian financial hub, following a similar step by Goldman Sachs. The decision signals growing caution among global lenders about third-party AI tools in markets where data protection, banking secrecy, and regulatory scrutiny are high.
The restriction applies to internal networks and devices used for work. It is aimed at reducing the chance of sensitive information entering external AI systems. The shift adds pressure on technology teams to provide safe alternatives and clear rules for employees who rely on AI to draft documents, summarize research, and code.
Why Banks Are Tightening Access
Large banks have been wary of consumer AI tools since early 2023, when staff began testing chatbots for daily tasks. Compliance officers raised alarms about confidential data leaving the firm’s walls. Model outputs that mix public and private information also pose legal and reputational risks if they contain errors.
Common worries include client privacy, market abuse rules, model bias, and the accuracy of generated content. Many firms have set up internal guardrails, such as disabling copy-paste of client details into public tools and logging prompts for review. Some are steering staff to in-house systems run on private clouds with audit trails.
What The Move Says About Claude
Bank follows Goldman Sachs in preventing use of Claude in Asian financial hub
Claude, created by Anthropic, has been popular for drafting, summarizing, and coding. It is known for its safety-focused design. But banks say even safety-minded tools can create risk if employees feed them sensitive inputs. The latest restriction suggests that brand or model design is less important than the firm’s control over data flows.
The bank’s action mirrors policies many institutions apply to a range of AI products, from chatbots to code assistants. The focus is on where data goes, whether prompts are logged by vendors, and if responses can be explained and tested.
Regulatory Pressures In Asia
Asian financial centers have ramped up governance expectations on AI and data analytics. Supervisors in major hubs ask banks to show clear accountability, strong model testing, and data protection. Firms must document who owns the risk, how systems are monitored, and how customer information is shielded.
That pressure makes public AI services hard to use at scale. Banks need contracts that define data handling, storage location, and breach processes. Public web tools often lack those guarantees. As a result, companies restrict access while they build vetted solutions.
Impact On Staff And Operations
For employees, the change could slow certain tasks, at least in the short term. Teams that used AI to draft research notes or write code will need approved workarounds. Technology leaders often respond by offering controlled pilots, with:
- Private instances of large language models hosted on secure infrastructure.
- Redaction tools to scrub sensitive data before prompts are sent.
- Human-in-the-loop reviews for high-impact outputs.
- Training on safe prompt design and documentation.
These steps aim to preserve efficiency gains without risking a data leak. They also create evidence trails for audits and regulatory reviews.
Market And Client Considerations
Clients expect banks to protect their information while still innovating. That means careful selection of vendors and strict testing. Restricting public tools can reassure clients that their data will not be exposed. But if banks move too slowly, they may fall behind peers that deploy secure AI at scale.
Analysts say the most successful firms will separate experimentation from production. Sandbox projects can explore new features, while critical operations rely on vetted systems with clear oversight.
What Comes Next
The latest restriction points to a broader trend: banks want AI, but on their terms. Expect more firms to tighten public tool access while investing in private models, data loss prevention, and policy training. Vendors that can offer strong privacy controls, clear audit logs, and on-premises options are likely to gain traction.
For now, the message is restraint. Institutions will keep testing AI under strict controls and prove that outputs are safe and reliable. The key question is how fast they can deliver secure, high-quality tools that match the ease of public chatbots—without putting client trust at risk.
