Generative AI (GenAI) has grown significantly, especially in Singapore. With the growth of the technology, the Infocomm Media Development Authority (IMDA) and AI Verify Foundation (AIVF) are working together to initiate regulation for GenAI in Singapore.
The launch of the insights from the Global AI Assurance Pilot and the introduction of the new Testing Starter Kit is a form of development that is happening in Singapore. Therefore, the regulation for GenAI in Singapore is very important.
The purpose of initiating regulation for GenAI in Singapore is to formalize norms for the implementation and testing of secure GenAI applications across the industry.
Also Read
Table of Contents
Background to the Initiation of Regulation for GenAI in Singapore
Since its launch in February 2025, the Global AI Assurance Pilot, designed to establish best practices for the technical evaluation of GenAI tools, has experienced numerous findings by IMDA of usage risks.
The program, which partnered 16 AI testers with 17 companies from 10 sectors including finance, healthcare, and public services, concluded that GenAI risks are indeed highly context-dependent. Factors that affect risk can manifest as use cases, industry, culture, and language, among others.
To address this, IMDA and AIVF initiated that this pilot project requires continuous collaboration with subject matter experts to determine regulation for GenAI in Singapore. The pilot project will continue to be refined based on feedback and sectoral needs.
Testing Starter Kit for GenAI Applications
Based on the findings from the Global AI Assurance Pilot, IMDA is introducing the Testing Starter Kit as the first step towards establishing regulation for GenAI in Singapore. Currently, the sedan testing kit is in the open process for industry feedback.
As a regulatory reference for GenAI in Singapore, there are currently four key risk categories from the evaluation of GenAI systems by the testing kit. Here are the four main risk categories:
- Hallucination (creation of inaccurate content)
- Unintended content (inappropriate or malicious output)
- Data disclosure (exposure of sensitive information)
- Vulnerability to unintended commands (manipulation through input engineering)
These risks guide the testing steps for use across all industries. This guide provides a platform for companies to carry out the recommended tests, through the completion of tools such as Project Moonshot. The kit will continue to evolve as feedback is received and technical advances are made.
Conclusion
Since the launch of the Global AI Assurance Pilot, IMDA has found several risks in its use. Factors that influence risk can include use cases, industry, culture, and language.
To address this, IMDA and AIVF initiated that the pilot project requires continuous collaboration with subject matter experts to determine regulation for GenAI in Singapore.
As a first step in addressing the imminent risks of GenAI, a Testing Starter Kit was introduced. The four main risk categories of GenAI system evaluation are hallucination, unwanted content, data disclosure, and vulnerability to unwanted commands.
These risks guide the testing steps for use across all industries and will continue to evolve as feedback is received and technical advances are made.
To find out the regulatory development process for GenAI in Singapore, it is better to work with Type Approval Certification Services for ICT Products. That way you will get updates regarding all information on the development of IMDA regulations in Singapore.
With experience and a solid team, Dimulti as Type Approval Certification Services for ICT Products can assist your company in facilitating Type Approval arrangements for IMDA certification in Singapore.
