AI Regulation Tracker / Guidance and frameworks
Singapore IMDA Issues Voluntary GenAI Chatbot Info-Card Guide
On July 20, 2026, Singapore's Infocomm Media Development Authority (IMDA) published voluntary Transparency Guidelines for GenAI Chatbots. At the center is a "chatbot info card," a consolidated, plain-language disclosure that IMDA compares to a nutrition or medicine label, covering what a chatbot can and cannot do, how reliable and safe it is, how data is handled, and how to report issues. Named adopters include Google's Gemini app, Meta, DBS, OCBC, and Singapore Airlines. The guidelines are voluntary and are not a binding rule.
Read the framing carefully, because it is the point. IMDA is not regulating the underlying models. It encourages the organisations that deploy chatbots to consumers to be clear with users about the specific application in front of them. In IMDA's words, the guidelines aim to "shift the conversation from model-level to application-level transparency." That is a deliberate choice. The people using a banking assistant or an airline helper do not need a research paper on the base model. They need to know what this particular chatbot does, where it fails, and what happens to what they type.
What the info card is
The centerpiece is a chatbot info card. IMDA says it can take different forms, such as a dedicated information page or a disclosure document, and it compares it to a nutrition label or medicine label, a single consolidated place to find the essentials about the chatbot you are using. The card is built on three principles that are easy to remember. Relevance means it should answer what users actually care about, namely what the chatbot can and cannot do, how reliable and safe it is, how their data is handled, and how to raise concerns or report issues. Accessibility means it should be in plain language, easy to read through, and easy to find. Timeliness means it should stay current, published when the chatbot goes live and updated when significant changes affect its capabilities, risks, or safety policies.
The guidelines themselves set three aims. They want to enable informed use, so users understand a chatbot's capabilities and limitations and can verify outputs when appropriate. They want to standardise disclosure, so users can compare chatbots without hunting through scattered documentation. And they want to enhance accountability, so deployers state their own safety commitments in public and can be assessed against them.
Why voluntary still matters here
It would be a mistake to file this under optional and move on. The adopter list is what gives a voluntary framework its weight. IMDA says organisations across the AI ecosystem have expressed support and that several indicated the guidelines will be a reference point over the next 12 months as they improve transparency for their public-facing chatbots. The named organisations include global AI companies such as Google, for the Gemini app, and Meta, along with major Singapore deployers such as DBS, OCBC, Singapore Airlines, and Synapxe. When that set of names commits to a common disclosure format, it becomes the standard that others get compared against, even without a statute behind it.
Sector regulators are lining up alongside it rather than replacing it. IMDA notes that the Monetary Authority of Singapore and the Ministry of Health have issued their own AI guidance for finance and healthcare, both emphasising transparency, and that the chatbot guidelines complement that sector-specific guidance where organisations deploy GenAI chatbots. Public agencies are signalling adoption too, with the National Library Board and the Health Promotion Board planning to take reference from the guidelines for their own chatbots. That combination, private deployers plus sector regulators plus the public sector, is how a voluntary framework hardens into an expectation.
Why this reaches US practice
IMDA regulates in Singapore, and these guidelines impose nothing on anyone in the United States. They reach US professionals because they hand you a clean, portable template for a disclosure that customers, boards, and regulators are increasingly going to ask for. If you run a consumer-facing chatbot, the info card is a practical checklist. Can you state, in plain language, what your chatbot does and does not do, how reliable and safe it is, how you handle user data, and how someone reports a problem? Most teams cannot produce that on demand today, and the exercise of writing it surfaces gaps worth fixing.
The application-level framing is the strategic takeaway. Much of the AI-governance conversation has focused on the model, but the entity your customer actually interacts with, and the entity that carries reputational and legal exposure, is the deployed application. A single well-maintained info card per chatbot is a low-cost way to show you understand that distinction. It is voluntary in Singapore, but as a discipline it travels. For US product and compliance leaders, building the info card now is cheaper than assembling one under pressure when a customer, an enterprise buyer, or a regulator asks.
Questions professionals are asking
Are these guidelines mandatory?
No. IMDA describes them as voluntary. They do not require or ban anything and do not create legal liability. They are a practical framework that GenAI chatbot deployers may reference and adopt, and IMDA says several organisations will use them as a reference point over the next 12 months.
What is the chatbot info card?
It is a consolidated, plain-language disclosure about a specific chatbot, which IMDA compares to a nutrition or medicine label. It can be a dedicated page or a disclosure document, and it should cover what the chatbot can and cannot do, how reliable and safe it is, how data is handled, and how to report issues. IMDA judges it on relevance, accessibility, and timeliness.
Who has said they will use it?
IMDA names global AI companies including Google, for the Gemini app, and Meta, along with major Singapore deployers such as DBS, OCBC, Singapore Airlines, and Synapxe. Public agencies including the National Library Board and the Health Promotion Board plan to take reference from the guidelines for their own chatbots.
Does this affect organisations outside Singapore?
Not as a rule. The guidelines apply in Singapore and bind no one elsewhere. For others they are a portable template. If you run a consumer-facing chatbot, the info card is a useful checklist for the disclosure that customers and regulators are increasingly likely to expect, and the application-level framing travels well.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any guideline, statute, or requirement applies to your situation with qualified professionals in the relevant jurisdiction.