- Singapore has issued new guidelines clarifying how the Personal Data Protection Act (PDPA) applies to personal data used in generative AI (GenAI) models and systems.
- The Personal Data Protection Commission (PDPC) guidelines specify that if no consent exception applies, organizations must obtain consent through AI-Specific Notifications.
- Singapore's Infocomm Media Development Authority (IMDA) has introduced voluntary transparency guidelines for generative AI chatbots, including a standardized 'chatbot info card'.
- The nutrition label guidelines for GenAI chatbots aim to provide users with essential information about the chatbot's capabilities and data practices.
- The PDPC guidelines cover web-scraping, the reuse of data originally provided for non-GenAI purposes, and the responsibilities of different GenAI stakeholders.
- Minister for Digital Development and Information Josephine Teo emphasized the importance of data governance in AI, stating that without good data, even the best systems will struggle to produce useful outcomes.
Singapore's Infocomm Media Development Authority (IMDA) has introduced new guidelines aimed at enhancing transparency in generative AI applications. The guidelines include mandatory AI-Specific Notifications for data training, requiring organizations to obtain consent from individuals when using their data.123
These notifications must clarify the purpose of data usage, the type of information involved, and how individuals can opt-out. Minister for Digital Development and Information Josephine Teo emphasized the importance of informed use, stating, “Without good data, even the best systems will struggle to produce useful outcomes.”6
In addition to the mandatory notifications, IMDA has also rolled out voluntary guidelines for generative AI chatbots, introducing a standardized “chatbot info card”. This card is designed to provide users with essential information about the chatbot's capabilities, limitations, and data practices, akin to a “nutrition label” for AI.4
The guidelines encourage companies to create infocards that are relevant, accessible, and timely, addressing user concerns about reliability, safety, and data handling. “The label does not tell us every scientific detail. Instead, it tells us the essentials,” said Teo.
Major companies like Google, DBS, and Singapore Airlines have indicated they will adopt these guidelines, which aim to strengthen accountability and support Singapore's broader push for trusted AI deployment.
The guidelines were developed following feedback from a public consultation that included responses from 40 organizations, highlighting the collaborative effort to enhance AI governance in Singapore.
“The Infocomm Media Development Authority (IMDA) said the voluntary chatbot infocards will be similar to medicine labels, providing essential information on capabilities, limitations, and data practices. Separately, the Personal Data Protection Commission (PDPC) clarified that organizations must obtain consent through AI-Specific Notifications when scraping personal data behind paywalls or without an exception.”
