Consultation: Model AI Governance Framework for Generative AI

ACCA welcomes the opportunity to comment on the proposed framework issued by IMDA Singapore and Verify Foundation, Singapore. We support the importance and need for a trusted eco-system for AI and commend this initiative particularly given the rapid developments in using generative AI globally. Within this eco-system, most of our members are either users of AI, and/or have a responsibility for the governance of AI systems – with a minority increasingly also upskilling in areas like programming. Therefore, our comments are principally from the perspective of deployers (as opposed to developers) and those conducting the assurance of AI systems. In forming our responses, we draw on research and policy work conducted over the years that involved surveys, interviews, and roundtables across a diverse base, including professional accountants in business, public sector, and public practice; as well as stakeholders in academia, industry bodies, and thought leaders in AI.

Overall, the proposed Framework for Generative AI is a useful step forward towards developing an eco-system for trusted AI. This is current, live issue as highlighted by ACCA’s 2024 Global Talent Trends survey of 9,889 finance professionals in 157 countries which reveals that 37% of respondents feel overwhelmed by the pace of change of technology impacting their job. The focus on Generative AI recognises the fast pace of change in AI with further considerations constantly entering the mix – for example we agree that issues like value alignment are even more in focus given the risk of unintended consequences from prompt engineering.

We agree with the nine dimensions noted in the Proposed Framework as being highly relevant for creating a trusted eco-system for Generative AI and at an appropriate level of detail that balances specificity with a high-level strategic view of the key considerations at stake.

We also agree that these dimensions need to be considered in an inter-connected manner in order to be meaningful to their fullest extent. The Content Provenance dimension for instance clearly links to the Data dimension on quality issues beyond the organisation’s own data sets.

We support a multi-stakeholder approach in line with our purpose to act as a force for public good. The AI for Public Good dimension within the framework is essential for this - citizens provide input data for AI models and are affected by the output decisions of model.

To read the response in full, please download the consultation document on this page.