Rapid corporate adoption of autonomous AI agents outpaces governance readiness in Singapore, necessitating urgent implementation of new state guidelines.
A significant majority of Singaporean businesses plan to deploy autonomous artificial intelligence (AI) agents within the next two years despite lacking mature oversight models, a report showed this week, highlighting the immediate challenge facing the city-state’s newly established regulatory framework.
The findings from Deloitte, released on Tuesday (3 February), reveal that 72 percent of organisations in Singapore intend to integrate “agentic” AI into their operations by 2028, a sharp increase from the 15 percent currently utilising the technology.
Advertisments
However, the consultancy warned that governance has not kept pace with ambition; only 14 percent of leaders reported having a mature framework to manage these autonomous systems, a figure falling below the global average of 21 percent.
This disparity highlights the critical timing of the Singapore government’s Model AI Governance Framework for Agentic AI, introduced on 22 January by Minister for Digital Development and Information Josephine Teo at the World Economic Forum in Davos.
The framework, developed by the Singapore’s Infocomm Media Development Authority (IMDA), is the first in the world specifically designed to address the risks posed by AI software capable of independent reasoning and action.
Unlike generative AI, which typically provides content or recommendations for human review, agentic AI can execute tasks autonomously, such as processing payments or updating sensitive customer databases.
While this promises to automate repetitive enterprise productivity tasks and customer services, it introduces the risk of unauthorised or erroneous actions occurring without direct human intervention.
“As AI agents take actions directly… agentic AI requires new approaches to governance,” Deloitte noted in its report. The firm found that half of the respondents in Singapore currently rely on a mix of internal and public frameworks to assess risk, emphasising the need for the structured guidance provided by the new government model.
The private sector has begun mobilising to bridge this readiness gap. On 2 February, cybersecurity firm Armor Defense Inc announced a five-country initiative covering Singapore, Thailand, Malaysia, Indonesia, and the Philippines to assist enterprises in operationalising the new requirements.
“Singapore’s Model AI Governance Framework for Agentic AI (MGF) recognises what we have been telling clients: AI agents that can act autonomously need the same security rigor as any privileged user,” said Chris Drake, Founder and CEO of Armor.
The government’s framework mandates that organisations maintain “meaningful human control” over these systems. It outlines four key dimensions for responsible deployment: assessing and bounding risks by limiting agent autonomy; defining specific checkpoints requiring human approval; implementing technical controls such as baseline testing; and ensuring end-user responsibility through transparency.
Beyond software agents, the industrial landscape is also shifting. The Deloitte report indicated that 84 percent of Singapore businesses expect to utilise “physical AI”, systems that drive actions through machines or robotics, within two years.
Additionally, 77 percent of businesses cited data residency and “sovereign AI” as critical to their strategic planning, reflecting growing concerns over reliance on foreign-owned technologies.
Singapore aims to use these domestic initiatives to lead regional standards, currently heading the ASEAN Working Group on AI Governance to foster a trusted ecosystem across Southeast Asia.