Hang Seng Bank ("Hang Seng") announces that two of its pilot use cases have been selected for the Generative Artificial Intelligence (GenA.I.) Sandbox++ jointly launched by the Hong Kong Monetary Authority (HKMA), the Securities and Futures Commission, the Insurance Authority and the Mandatory Provident Fund Schemes Authority, with support from the Hong Kong Cyberport Management Company Limited.

The two pilots aim to improve customer experience and operational efficiency while strengthening safeguards for A.I.-related risks. A key feature is testing "A.I. vs. A.I." controls, using one A.I. system to review, challenge and validate another A.I.'s outputs, to support stronger governance, auditability and human oversight.

For retail banking, Hang Seng will test an "AI Judge" solution that reviews GenA.I. chatbot responses for bank servicing enquiries before content is delivered to users. The solution is designed to safeguard customer-facing GenA.I. chatbot and deliver more accurate, better-toned and compliant responses to users.

For commercial banking, Hang Seng will pilot GenA.I. to support customer onboarding and due diligence by checking business authenticity and legitimacy across multiple dimensions. The pilot is designed to validate business information, detect early red flags and generate an auditable summary of findings for faster customer onboarding while keeping humans in control of final decisions.

Both pilots will be conducted with robust customer data safeguards, including data minimisation, use of public or synthetic data and monitoring in a controlled environment.

Forrest Chai, chief information officer of Hang Seng Bank and HK, said: "Being selected again for the GenA.I. Sandbox++ reinforces Hang Seng's commitment to responsible innovation in Hong Kong. Trust is the new currency, and our strategic focus is to scale GenA.I. in ways that are safe by design—strengthening governance, auditability and human oversight from day one. Building on our first-cohort progress, we are advancing 'A.I. vs. A.I.' strategies to challenge and validate outputs, improving customer experience and operational efficiency while managing A.I.-related risks. We look forward to sharing practical learnings to support trusted GenA.I. adoption across Hong Kong's financial ecosystem."

Hang Seng's participation builds on its work in the first cohort of the GenA.I. Sandbox, where a proof-of-concept experiment on fraud investigation automation using synthetic data (with no real personally identifiable information (PII)), evaluation results on synthetic test cases found that over 90% of extracted information matched the reference. The Bank will continue to share practical lessons from responsible GenA.I. testing, with customer benefit and robust controls as the priority.

Re-disseminated by The Asian Banker