Singapore AI Safety Institute

The Singapore AI Safety Institute (AISI) is Singapore's designated national body for AI safety research, evaluation, and governance. It was formally established on May 22, 2024, when IMDA designated the Digital Trust Centre (DTC) at Nanyang Technological University (NTU) as the country's AI safety institute. The DTC itself had been created earlier in June 2022 with a S$50 million grant from IMDA and Singapore's National Research Foundation (NRF), led by Professor Lam Kwok Yan as Executive Director.

The institute's mandate is to address gaps in global AI safety science by leveraging Singapore's existing capabilities in AI evaluation and testing. Organizationally, IMDA handles baseline policy development and international engagements, while the DTC drives technical research and R&D efforts from its NTU base.

The AISI has four core work areas: testing and evaluation of AI systems, safe model design and deployment, content assurance, and governance and policy. Research programs include preventing harmful content from large language models, LLM robustness through adversarial testing, distinguishing AI-generated from human-generated content, and comprehensive safety testing frameworks.

The institute has been active in international collaboration. It joined the International Network of AI Safety Institutes, signed an MOU with Korea's AISI, and conducted a joint multilingual testing exercise with Japan covering 10 languages and five harm categories. In November and December 2024, IMDA and the AISI co-organized the world's first multicultural and multilingual AI Safety Red Teaming Challenge for Asia-Pacific, involving over 350 participants from nine countries testing four large language models for cultural bias.

In April 2025, Singapore hosted the Singapore Conference on AI: International Scientific Exchange on AI Safety, bringing together over 100 leading AI scientists from 11 countries including Yoshua Bengio and Max Tegmark. This produced the Singapore Consensus on Global AI Safety Research Priorities, organizing AI safety research into three domains: creating trustworthy AI systems (Development), evaluating AI systems' risks (Assessment), and monitoring and intervening after deployment (Control). In February 2025, Singapore also launched the Global AI Assurance Pilot in partnership with the AI Verify Foundation to codify best practices for technical testing of generative AI applications.