The Hong Kong Monetary Authority and Cyberport announced on Wednesday the second batch of the generative artificial intelligence sandbox to trial secure use cases, including pioneering "AI versus AI."
The second cohort, receiving over 60 proposals, picked 27 use cases from 20 banks and 14 technology partners, according to a statement by the city's de-facto central bank
Among them are banking giants like Bank of China (Hong Kong) (2388) and Hongkong and Shanghai Banking Corporation, digital banks like PAO Bank, and Chinese tech leaders including Alibaba Cloud and China Mobile Hong Kong.
With an aim to experiment with AI's capability to ensure its secure and reliable implementation, the participants were expected to explore proactive approach to AI governance.
Some use cases will trial using AI against AI, with an application scenario of leveraging the technology to conduct automated checks on AI-generated outputs, in a bid to improve accuracy and consistency at a scalable level, said HKMA in the statement.
To mitigate the emerging risks of deepfake-related fraud, the Sandbox also took in some adversarial simulations, using AI to stress-test and cement the participants' systems.
Besides, there are some proposals focusing on risk management, including adopting multi-agent AI for credit portfolio management, intelligent trade finance facilitation, and market risk monitoring and early alert.
The trial is expected to start in early 2026.
