The Singapore government introduced three new AI governance initiatives to promote safety and global best practices. The initiatives include the Global AI Assurance Pilot, which focuses on testing generative AI applications; a joint testing report with Japan to enhance AI safety across different linguistic environments; and the publication of the Singapore AI Safety Red Teaming Challenge evaluation report, aimed at addressing AI performance across languages and cultures.
The announcement was made by Josephine Teo, Singapore’s Minister for Digital Development and Information, at the AI Action Summit (AIAS) in Paris. During her speech, Minister Teo emphasised Singapore’s commitment to fostering international collaboration on AI safety, noting the importance of understanding public concerns and ensuring AI systems are tested for safety and responsibility. She also highlighted the role of private sector partnerships in shaping AI use cases and risk management strategies.
The new initiatives include practical efforts to ensure AI models, particularly large language models (LLMs), are secure and culturally sensitive. The AI Assurance Pilot, for instance, will bring together global AI assurance vendors and companies deploying real-life GenAI applications to establish future standards for AI governance. The joint testing report with Japan aims to improve the safety of LLMs across multiple languages, addressing potential gaps in non-English safeguards. Additionally, the Red Teaming Challenge provided insights into AI performance and cultural bias, with participants testing LLMs for issues such as violent crime and privacy violations.
For more information on these topics, visit diplomacy.edu.
