Singapore launches 'nutrition label' guidelines for GenAI chatbots

Singapore has introduced voluntary guidelines aimed at increasing transparency around generative artificial intelligence (GenAI) chatbots, encouraging companies to provide users with consolidated information similar to food packaging labels. The initiative, launched on July 20, marks a step toward clearer communication about the capabilities, limitations, and data-handling practices of AI-driven conversational interfaces. Minister for Digital Development and Information Josephine Teo announced the guidelines during the Singapore Data Festival, emphasizing the growing integration of GenAI into daily life. She noted that while users interact frequently with chatbots, many remain unaware of their constraints or how their data might be used. To address this, the Infocomm Media Development Authority (IMDA) proposed the creation of “infocards”, documents that would summarize critical aspects of chatbot functionality in an easily digestible format. These could appear as standalone pages or integrated within user interfaces, offering clarity on what the chatbot can and cannot do, how reliable it is, and how user data is managed. The infocards aim to consolidate information currently dispersed across lengthy terms of service and privacy policies, which often prove difficult for average users to comprehend. According to Teo, the goal is to mirror the simplicity of a medicine label, providing essential details without overwhelming users with excessive technicalities. This includes outlining the purpose of the chatbot, recommended usage, potential risks, and procedures for reporting issues. The guidelines stress the importance of keeping such information current and accessible, ensuring users receive accurate and up-to-date insights. Several organizations have expressed interest in implementing the guidelines, with the public sector leading the way. Agencies such as the National Library Board and Health Promotion Board plan to adopt the framework, setting a precedent for private-sector entities. Major corporations including DBS, Google, Meta, OCBC, and Singapore Airlines have also committed to integrating similar disclosures into their platforms over the coming year. While the guidelines are initially voluntary, the government expects widespread adoption as best practices evolve. In parallel, the Personal Data Protection Commission (PDPC) issued advisory guidelines requiring organizations to notify users when personal data is used to train GenAI models. Effective from July 20, these AI-specific notifications apply to entities collecting data for model development, such as call centers using recordings to enhance customer service algorithms. The PDPC clarified that while notification formats are not strictly defined, companies are encouraged to use in-app pop-ups or dedicated webpages to explain the nature and scope of data collection. Consumers retain the right to opt out, although refusal to participate does not justify denial of services. The PDPC emphasized that personal data, encompassing everything from contact details to biometric information, must be handled responsibly. Organizations may legally scrape publicly available data without explicit consent, but they must determine whether such data remains accessible under digital restrictions like paywalls or login requirements. If not, prior consent becomes necessary. The guidelines also clarify that existing privacy notices, previously sufficient for activities like product development, now require specific adaptation to cover AI-related data use. As both regulatory bodies push for greater transparency, the focus remains on balancing innovation with user protection. With GenAI becoming increasingly embedded in everyday interactions, the emphasis on clear, accessible information reflects a broader effort to ensure trust and informed decision-making among users. The next phase involves refining these frameworks based on industry feedback, ensuring they remain effective as technology continues to advance.