The Personal Data Protection Commission (PDPC) (Singapore's data protection regulator) issued the Model Artificial Intelligence (AI) Governance Framework (Model Framework) in January 2019.
The Model Framework does not impose any binding requirements on organizations. Instead, the Model Framework provides a compliance and governance framework that organizations who adopt and deploy AI solutions are encouraged to adopt. For observers in this area, the Model Framework also provides insight into potential future regulatory touchpoints and considerations.
The 4 key areas of governance proposed in the Model Framework are summarized below:
1. Internal governance structures and measures
Organizations should put in place internal governance structures and measures to ensure robust oversight over AI used by the organization. Governance structures and measures may be adapted from existing internal governance structures, otherwise new structures may be implemented if necessary. Clear roles and responsibilities for the ethical deployment of AI should be established and allocated. A sound system of risk management and controls should also be implemented.
2. Risk management in autonomous decision-making
Organizations should weigh the commercial objectives of using AI against the risks before deployment of such technologies. Due consideration should be taken to the differences in societal norms and values as well as the organization's corporate values when designing and implementing AI solutions. Organizations should continually identify and review risks relevant to their technology solutions, mitigate those risks, and maintain a response plan should mitigation fail.
3. Operations management
Organizations should be aware of the interaction between data, algorithms and the AI model when deploying an AI solution. Relevant departments in an organization should work together to implement good data accountability practices that allow the organization to understand the lineage of the data utilized both during development and deployment of AI. Factors that may affect the quality of the data should be understood and addressed. And there should be an awareness of and steps taken to mitigate any inherent bias that may be present in the dataset.
4. Customer relationship management
Organizations are encouraged to maintain appropriate communication channels with the public that inspire and build trust in the AI solutions utilized by the organization. Proper disclosure and transparency can assist the public in better understanding AI solutions deployed and their 1 implications. The human usability of an AI system should be considered before deployment as well as the ability to provide opt-out and channels for feedback and review of decision made by AI.
The PDPC expects the use of AI by businesses to continue to evolve, and intends to adapt the Model Framework accordingly. Organizations can share any feedback on the Model Framework with the PDPC by 30 June 2019 at email@example.com