Singapore urges AI chatbot providers to display 'information cards'

Singapore on Monday urged providers of generative artificial intelligence (AI) chatbots to present essential information in simple language through ‘chatbot information cards’, a core component of the newly launched voluntary Transparency Guidelines for Generative AI Chatbots. The guidelines, unveiled at the Singapore Data Festival by Minister for Digital Development and Information Josephine Teo, recommend that chatbot providers clearly state the chatbot’s capabilities and limitations, security and reliability practices, data use and protection, and channels for reporting complaints. Teo said the information users need is often presented in a fragmented manner. She likened the information card to a drug label, providing a single, centralised place where users can easily find important information about a chatbot. The information card can be provided in various formats, such as a dedicated webpage or a specific explanatory document, but must be written in simple, easy-to-understand and accessible language. Chatbot providers are also encouraged to update the information whenever significant changes occur that affect the chatbot’s capabilities, risks or security policies. According to a statement from Singapore’s Infocomm Media Development Authority, several organisations, including global AI companies Google and Meta, as well as major enterprises deploying AI in Singapore such as DBS, OCBC, Singapore Airlines and Synapxe, have indicated they intend to use the guidelines as a reference over the next six to twelve months. The statement noted that the guidelines currently focus on generative AI chatbots due to their scale, high level of consumer interaction, and growing public concerns regarding data privacy, the safety of minors, and risks to mentally vulnerable users. At the opening of the five-day Singapore Data Festival, the Singapore government also launched several other initiatives, including a practical guide to help companies adopt digital twin technology, and advisory guidelines on the use of personal data in generative AI.