Singapore Unveils ‘Nutrition‑Label’ Guidelines for GenAI Chatbots

When Singapore’s Data Festival opened its doors this week, a modest but notable announcement stole the spotlight: the Infocomm Media Development Authority (IMDA) is offering a fresh, voluntary “nutrition‑label” style infocard for every generative‑AI chatbot out there.

Minister Josephine Teo, who heads the Digital Development and Information portfolio, explained that the idea is simple – give users one clear place to see what a chatbot can actually do, the limits it has, how safe it is, and exactly how their data might be handled. Think of it as the food‑label you see on a snack, but for AI.

The guidelines, which IMDA says should be written in plain language, must be easy to find and easy to navigate. They need to list things like the chatbot’s core capabilities, any known reliability concerns, the kinds of personal data it may collect, and a quick route for users to raise complaints or seek help.

While the framework is voluntary, officials are nudging organisations to adopt it within the next six to twelve months. So far, a handful of public‑sector bodies – the National Library Board and the Health Promotion Board – have already pledged to roll out the infocards. On the private side, big names such as DBS, Google, Meta, OCBC and Singapore Airlines have also signalled their intention to join in.

Alongside the label, IMDA released another set of advisory guidelines aimed at responsible data collection for training GenAI models. The twin moves are meant to boost trust, especially as chatbots become more embedded in everyday services.

Nothing in the announcement spells out penalties for non‑adoption; the approach leans on goodwill and the hope that clear, user‑first information will become the norm rather than the exception.