Singapore introduces voluntary disclosure guidelines for generative AI chatbots

Singapore’s Infocomm Media Development Authority (IMDA) has introduced voluntary guidelines outlining how organisations should explain the capabilities, limitations and safeguards of public-facing generative AI chatbots.

The Transparency Guidelines for GenAI Chatbots aim to provide users with a clearer understanding of an application’s capabilities, reliability, safety measures, data handling practices and reporting channels. By encouraging a more consistent format, the IMDA intends to help users better assess whether a chatbot fits their specific needs and identify when its responses require independent verification.

Information cards bring disclosures into one place

The framework centres on a chatbot information card, which can be implemented as a dedicated webpage or a digital disclosure document. The IMDA compared this approach to nutrition or medicine labels, which consolidate all necessary assessment details into a single, accessible location.

Each card should outline the chatbot’s capabilities and limitations, the measures used to manage safety and reliability, data handling practices and channels for reporting issues. This information must be written in plain language, easily located and straightforward to navigate.

Maintaining up-to-date information is a core component of the framework. Providers are encouraged to publish the card alongside the chatbot’s launch and update it whenever changes alter the application’s capabilities, risks or safety policies.

Collectively, these requirements establish a more consistent standard for disclosure across different services. While providers publicly declare their commitments, users gain a single reference point to understand how a chatbot is designed and managed.

The IMDA developed these guidelines following consultations with technology firms, operators of consumer-facing chatbots, public agencies and local users. The framework focuses on delivering practical consumer insights while allowing providers operational flexibility in how they present the information.

Adoption will begin across companies and public agencies

Several organisations intend to use the framework to review their current disclosure practices. Major entities including Google, DBS, OCBC and Singapore Airlines have indicated plans to enhance transparency for their public-facing chatbots over the next six to 12 months.

In the public sector, the National Library Board, Health Promotion Board and Land Transport Authority also plan to align their chatbot services with the new guidelines.

While these commitments signal early adoption, individual organisations retain full flexibility over implementation. Consequently, the long-term consistency of these disclosures will depend on how each provider applies the guidance and updates the information as their services evolve.

The new framework will operate alongside existing regulations in specific sectors. For instance, the Monetary Authority of Singapore’s proposed guidelines on AI risk management and the Ministry of Health’s updated Artificial Intelligence in Healthcare Guidelines both emphasise transparency as a pillar of responsible AI deployment.

The framework may extend to other AI applications

The IMDA prioritised chatbots due to their direct, widespread interaction with consumers and the associated risks regarding data privacy, the safety of minors and protection for mentally vulnerable users.

However, these core disclosure principles could eventually be extended to other generative AI applications. Sector-specific regulators may also adapt the framework to suit industries like healthcare, education and financial services, where user information needs vary based on distinct risk profiles.

Moving forward, the IMDA plans to continue collaborating with regulators and industry stakeholders as these applications become more deeply integrated into daily operations.