Amid worries of growing AI risk, the Singapore government has unveiled a governance framework that seeks to guide organisations to safely deploy and use increasingly smart and autonomous AI agents.
The first such framework in the world, it recommends both technical and non-technical measures to mitigate risks, while emphasising the importance of humans in the loop to be ultimately accountable.
Singapore’s infocomm regulator, which developed the framework, says it supports responsible AI, so that the technology’s benefits can be enjoyed in a trusted and safe manner.
The country’s practical and balanced approach puts in place guardrails, while providing space for innovation, according to the Infocomm Media Development Authority (IMDA) today.
Its Model AI Governance Framework for Agentic AI, launched at the World Economic Forum in Davos, builds on earlier efforts by Singapore to lead the way in AI development and usage.
In 2020, it had put out a governance framework for AI in general. Since then, AI has grown beyond chatbots, with agentic AI now promising to reason and take action autonomously on behalf of human users.
The worry is that these AI agents will make serious errors, for example, making unauthorised payments, or exposing the sensitive data they have access to.
Singapore’s framework advises organisations to select the right use cases for AI agents and limit their autonomy and access to tools and data. It also calls on humans to be made accountable for the agents by requiring approvals for certain actions.
Technical controls through an AI agent’s lifecycle and end-user training and education are other suggestions that Singapore recommends for safe agentic AI deployment.
The framework is a “living document” that will be updated as the technology progresses rapidly in the coming months and years. Feedback from the industry is key as well, says IMDA.
However, this latest effort doesn’t come with regulatory requirements for organisations to follow. Unlike the European Union, which has some of the toughest regulations on AI so far, Singapore has mainly relied on existing laws such as ones for personal data protection to manage the risks of AI to the public.
Despite that, various regulators in the country have proposed more specific guidelines for certain industries in recent months, as it bets big on the nascent technology.
The Cyber Security Agency has put out a draft advisory on securing AI systems, while the Monetary Authority of Singapore has proposed AI oversight and risk management guidelines for the financial sector.
Speaking in Davos this week, Minister for Digital Development and Information, Josephine Teo said it was key to preserve human choice while harnessing agentic AI.
“As Singapore explores using AI agents to make government services more accessible, we’ll start with low-risk transactions to build trust whilst providing physical touchpoint options,” she noted, in a Facebook post today.
Gaining capabilities also helps to achieve technological sovereignty, she added. “Singapore accesses the best global technologies and maintains sovereignty by setting standards and assessing compliance.”
