The Hong Kong Monetary Authority ( HKMA ), in collaboration with Hong Kong Cyberport Management, has launched the second cohort of the generative artificial intelligence ( GenAI ) sandbox.

A total of 20 banks and 14 technology partners will take part in the initiative, which is positioned as a cornerstone in Hong Kong’s broader strategy to promote responsible AI adoption in finance.

The first cohort, launched on December 19 2024, focused primarily on demonstrating how GenAI could enhance risk management, anti-fraud measures, and overall customer experience. It showcased AI’s strength in processing large volumes of unstructured data – including documents, text, and multimedia – and improving credit assessments, fraud detection, and personalized customer servicing.

A defining feature of the second cohort is its focus on AI governance and AI assurance. Instead of simply embedding AI into workflows, participating banks will test "AI vs AI" models – leveraging machine learning systems to monitor, evaluate, and validate the performance of other AI tools.

This approach aims to address one of the industry’s most pressing challenges: maintaining accuracy, consistency, and trustworthiness in AI-driven decisions. With these automated quality-assurance mechanisms, banks can achieve scalable compliance while avoiding model drift and unintended bias, according to the HKMA.

Adversarial simulations

Another key focus is deepfake-related fraud prevention, which has become increasingly relevant across global financial systems. Financial fraud reports for 2024 indicate a significant rise in digital and identity-related crimes, with synthetic identities and deepfakes being used to carry out sophisticated attacks, particularly in the financial services sector.

Several participants will engage in adversarial simulations, deploying AI both as a defensive and offensive mechanism to stress-test their detection infrastructure against sophisticated digital deception tactics.

PAObank and OneConnect Financial Technology will work closely within the sandbox to refine deepfake detection and verification systems. These efforts are expected to enhance authentication processes in remote onboarding, video verification, and digital identity checks – all within a controlled, risk-mitigated environment.

By reinforcing governance and risk frameworks around GenAI, Hong Kong is positioning itself as a regional leader in AI-driven financial innovation, the HKMA says. This approach balances two priorities: encouraging creative experimentation and ensuring robust regulatory oversight.

These initiatives align with Hong Kong’s Fintech 2025 strategy, which seeks to modernize the financial sector through technology adoption, regulatory clarity, and upskilling. The sandbox, in particular, provides a testbed for both local banks and international institutions exploring the intersection of innovation and compliance.

“We are pleased to see a keen interest from the banking industry in the exploration of GenAI, reflecting the eagerness and openness of banks of all sizes to utilize novel technologies in their operations and services,” says HKMA deputy chief executive Arthur Yuen. “As the market continues to develop more innovative ideas, the GenAI Sandbox remains ready to provide a risk-controlled testing ground for the industry.”