The concerns come amid a push by the Hong Kong Monetary Authority to facilitate the wider adoption of regulatory technology, or regtech. Arthur Yuen Kwok-hang, deputy chief executive of the HKMA, Hong Kong’s de facto central bank, introduced four regtech initiatives last week during the annual conference of the Hong Kong Institute of Bankers.
The increased use of AI and machine learning creates challenges for banks, wherein they must account for decisions made by these technologies. The use of regtech could reduce their ability to be accountable for the decisions made.
Guy Sheppard, head of APAC financial crimes, analytics and intelligence at Swift, a network that enables financial institutions worldwide to send and receive information about transactions in a standardised environment, said the outcome of a bank’s internal model will dictate how it, for example, off-boards certain types of customers even before they have conducted any financial crimes.
“I would like to see how regulators will address a bank off-boarding a certain profile of customers after their internal model has dictated that they are beyond the bank’s stated risk appetite,” he said.
Deep learning, which uses sophisticated mathematical modelling to process data in complex ways, offers few clues on how it arrived at a conclusion. Sheppard said such lack of transparency was a “pain point” regulators would have to accept with regard to regtech.
In perhaps an admission of the challenges that lie ahead, Yuen told the conference: “It is not technology at any cost. We cannot outsource the responsibility to technology firms – we expect banks to understand why a machine makes the recommendations it does.”
To help develop a regtech ecosystem, the HKMA will open a fintech “supervisory ‘sandbox’ to regtech projects or ideas raised by banks and technology firms”, he said, adding that this was likely to be effective this week.
The sandbox – a highly controlled environment used to test unverified ideas – will allow the banks and their technology partners to conduct pilot trials of newly developed technology, without the need for full compliance.
The development and use of regtech by banks to achieve regulatory compliance and automate risk management has grown in recent years. The use of regtech for surveillance, in particular, has grown rapidly as part of anti-money-laundering and anti-terror financing measures.
The HKMA initiatives will also spearhead the use of regtech for prudential risk management, as well as the use of technologies related to machine-readable regulations and supervisory processes and activities.
Yuen said the number of suspicious transactions reported by Hong Kong banks had grown on average by 40 per cent annually over the past five years. He said some banks were using machine learning and artificial intelligence to help detect suspicious behaviour and patterns, which had freed up analysts to focus on high-risk cases, thus enhancing the overall effectiveness and efficiency of banks’ transaction monitoring processes.
Benjamin Quinlan, chief executive and managing partner at consultancy Quinlan & Associates, said the industry did welcome the HKMA’s exploration of machine-readable regulations.
Machine-readable regulations help banks automate the interpretation of regulatory requirements. “[If regulators] provide two to three pages of guidelines, on what regulations mean … that would help a regtech provider understand how the regulations are going to work. The job of a regulator [is] to set the tone and make [regulations] as clear as possible,” said Quinlan.