Hong Kong issues generative AI guidelines for consumer protection

The Hong Kong Monetary Authority (HKMA) has introduced new guiding principles for the use of generative artificial intelligence (AI) in consumer-facing applications, aimed at protecting end-users. These principles build upon the guidelines that were released in November 2019 and relate to big data and AI. The changes include governance and accountability, and integrity standards, including fairness, transparency and disclosure, data privacy, and protection.

Alan Au, the executive director of HKMA’s banking conduct department, pointed to the increasing popularity of generative AI in treating banking customers. Some examples of applications include customer service chatbots, personalized product creation, sales, marketing, and even robo-advisors in the financial planning and insurance industry. As the use of AI is on the rise, the HKMA has emphasized the importance of accountability by demanding that institutions give consumers understandable notices.

Furthermore, the HKMA asserts that boards and senior management must bear full responsibility for decisions on generative AI. The models applied must produce objective and nondiscriminatory results for consumers.

HKMA and Cyberport Launch AI Sandbox for Safe Testing

At present, there are no laws or regulations uniquely governing generative AI in Hong Kong, but its regulators are proactively addressing the growing popularity of this technology. The Office of the Privacy Commissioner for Personal Data released the first batch of personal data protection recommendations for organizations applying generative AI. These guidelines suggest that an internal AI governance committee should be formed and that checks for compliance should be conducted periodically.

To enhance the accountability of AI usage, the HKMA, jointly with the government-backed incubator Cyberport, introduced a generative AI sandbox. This initiative offers financial institutions an opportunity to test AI applications within a constrained, risk-managed environment, especially given technical support. The sandbox is expected not only to assist institutions in addressing the difficulties and obstacles of introducing AI but also to protect consumers.