Why Singapore’s GenAI chatbot transparency rules may offer an APAC disclosure reference

Consumer-facing chatbots often ask users to accept a great deal of information without putting much of it in one place. What the tool does, its limits, safety measures and privacy details can sit across product pages, terms and notices.

Singapore’s Infocomm Media Development Authority (IMDA) has issued guidance for a more usable arrangement. Its Transparency Guidelines for Generative AI Chatbots, released on 20 July 2026, are voluntary, but they ask deployers to bring the core explanation together in a chatbot information card that users can find when they start using the tool.

What happened

The guidance is aimed at consumer-facing chatbots. IMDA strongly encourages deployers to provide at least one substantive disclosure in each of four information areas. It also recommends a high-level safety statement and an identifiable link to the information card at first use.

The three practical principles are relevance, accessibility and timeliness. In plain English, the information should help a person understand the chatbot they are using, be easy to locate without a hunt through legal pages and change when the product changes.

IMDA says Google, Meta, DBS, OCBC, Singapore Airlines and Synapxe supported or participated in developing the guidance. The National Library Board and Health Promotion Board plan to take reference from it for their public-facing chatbots. This is early institutional engagement with the guidance.

What it means

The card does not settle whether a chatbot is safe, accurate or legally compliant. It creates a public record of what its deployer says the tool is for, what safety information it provides and where users can find the fuller explanation.

In Relay’s reading, the card gives product, privacy, risk and customer teams a reason to keep one clear account of the chatbot. For example, if a bank changes a chatbot so it can access account data, the card should be updated to explain what changed, what data is involved and which safeguards apply. Otherwise, a customer may share information without knowing that the chatbot’s role has changed.

Looking at other APAC materials reviewed for this piece, Australia is the closest disclosure comparator: its privacy regulator says public-facing AI tools, including chatbots, should be clearly identified to external users. Japan’s voluntary guidance asks organisations to manage AI risks across the lifecycle, while South Korea’s statutory framework addresses trust and high-impact AI. These are broader governance layers, not direct equivalents to a consumer-chatbot disclosure card.

Against that backdrop, Singapore’s guidance may be the most specific source reviewed here on a dedicated consumer-chatbot disclosure method at the time of publication. Its distinction is practical: it turns a broad transparency principle into information a user can access at the moment of use.

Its influence will depend on uptake and on whether the card remains a useful explanation rather than another layer of corporate copy.

What to watch

• Published cards: Whether consumer-facing chatbot deployers publish identifiable cards and place the link where users first encounter the tool.

• Current information: Whether cards are revised after material product or policy changes, with enough detail to remain useful.

• Regional adaptation: Whether a market adopts a comparable consumer-facing disclosure method, rather than simply citing a broader AI-governance principle.