These initiatives aim to keep pace with the rapidly evolving technology and ensure that AI systems can be deployed safely. Mr. Tan highlighted that government agencies, including the Government Technology Agency of Singapore (GovTech), are involved in these trials with appropriate safeguards. He also encouraged private sector organizations to exercise care and responsibility when deploying AI agents.
Agentic AI systems, capable of executing actions, interacting with external systems, and adapting their behavior with reduced human oversight, present both opportunities and risks. Mr. Tan noted that many of the risks associated with these systems are extensions of existing challenges, and current guidelines regarding human accountability, controls, and cybersecurity remain applicable.
The government is also monitoring international developments in this emerging field and is developing internal capabilities to go beyond current frameworks. This includes a focus on strengthening cybersecurity to protect systems that AI agents interact with, as well as adapting frameworks to account for the increased autonomy of these systems.
Mr. Tan welcomed frontier AI companies to take greater responsibility in managing the risks of their systems, including early identification of capabilities and risks, and implementing safeguards before widespread deployment. He stressed the importance of rectifying issues promptly if AI systems cause harm or impact others.
Minister for Digital Development and Information Josephine Teo added that Singapore is studying the need for new safety measures in high-risk AI applications. Potential safeguards for essential services and other high-risk uses could include more rigorous testing, verifiable evidence of safety measure effectiveness, tighter deployment controls, and stronger regulatory oversight.