The Personal Data Protection Commission of Singapore (PDPC)
The Singapore Model AI Governance Framework (Model Framework) is a policy guidance framework created by the Personal Data Protection Commission of Singapore (PDPC). The Model Framework provides comprehensive and actionable guidance to private sector organisations for managing the ethical and governance aspects of deploying AI solutions. The Model Framework emphasises the importance of understandable AI systems, implementation of robust data accountability practices, and fostering open and transparent communication. With the primary objective of enhancing public trust and comprehension of AI technologies, the Model Framework was initially introduced by the PDPC in January 2019 and subsequently updated in January 2020. The Model Framework is accompanied by the Implementation and Self Assessment Guide for Organisations (ISAGO) that assists organisations in evaluating how closely their AI governance practices align with the Model Framework, while also offering a comprehensive repository of valuable industry examples and best practices to facilitate the adoption of the Model Framework.
The Model Framework is policy guidance directed for private-sector organisations. Organisations in the scope of the Model Framework cover companies and other entities that adopt or deploy AI solutions in their operations. The operations can include for example, backroom operations such as processing applications for loans, front-of-house services such as a ride-hailing app or e-commerce portal, or the distribution or sale or of devices that provide AI-powered features, such as smart home appliances.
The Model Framework operates on two overarching guiding principles, aiming to cultivate trust and comprehension in AI technology use: Firstly, it advocates for AI decisions to be explainable, transparent, and fair, and secondly, it emphasises the need for AI systems to be human-centric. From these guiding principles, the framework has derived its four primary domains that guide responsible and ethical AI deployment. The first one being establishing internal governance structures and measures that involve defining clear roles, implementing risk management procedures, and providing staff training. The second domain covers the degree of human involvement in AI-augmented decision-making that centers on finding the right level of human participation and minimising potential harm. The third domain concerns efficient operations management that aims to reduce bias, adopt a risk-based approach, and ensure explainability, robustness, and ongoing model tuning. The fourth and last domain consists of stakeholder interaction and communication that focuses on sharing AI policies, gathering user feedback, and ensuring clear and comprehensible communication.