peshkov -

Singapore updates model AI governance framework

Second edition of Singapore’s artificial intelligence governance framework includes new guidance, use cases and a self-assessment guide

The Singapore government has refined its model artificial intelligence (AI) governance framework to improve its relevance and usability for organisations that are implementing AI.

The second edition of the model framework includes examples of how organisations have implemented AI governance practices and guidance on the level of human involvement in AI-augmented decision-making, among other changes.

First unveiled a year ago, the model framework, one of the first of its kind, is underpinned by two high-level guiding principles – that AI implementations should be human-centric, and that decisions made or assisted by AI should be explainable, transparent and fair to consumers.

These principles will enhance trust in and understanding of AI, as well as acceptance of how AI-related decisions are made for the benefit of users, according to the Infocomm and Media Development Authority (IMDA).

Since its introduction, the framework has been adopted by more than 15 organisations of all sizes – from global financial firms such as DBS and HSBC, Mastercard and Visa, to local institutions including Singapore’s Ngee Ann Polytechnic.

Speaking at the World Economic Forum in Davos this week, Singapore’s minister for communications and information, S Iswaran, said the fact that international organisations have adopted the framework demonstrates its value in providing practical guidelines on how AI can be implemented while maintaining the trust of clients and customers.

To help organisations measure their progress in adopting the framework, the IMDA has also developed an implementation self-assessment guide, which Iswaran said will help organisations “calibrate how they are making progress”.

Also, a compendium of use cases will demonstrate how local and international organisations have implemented or aligned their AI governance practices with the model framework.

Singapore’s DBS Bank, for example, aligned the implementation of its anti-money laundering AI model to the framework and had considered, among other things, the probability-severity of harm matrix.

It also adopted the human-over-the-loop approach to give its AI model the freedom to decide which transactions were suspicious, while allowing humans to intervene when necessary.

Read more about AI in ASEAN

In using AI to improve admissions selections, Ngee Ann Polytechnic has also implemented internal governance structures and measures, such as final management approval of all AI deployments and manually reviewing admission applications that were not selected by its Eva admissions chatbot.

“Whether it is the model AI governance framework or the self-assessment guide or this compendium of use cases – ultimately, what we want is to engender trust in this technology,” said Iswaran, adding that the framework will benefit businesses by lowering barriers to AI adoption.

“You know what the technology is, the possible solutions – how do I implement it, how do I ensure that in the process of implementing, I maintain the confidence and trust of my consumers, my clients, my counterparts. This is an important aspect of any organisation’s efforts to implement AI, and in that regard, this will be very valuable,” said Iswaran.

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management