If you run a Malaysian business that has started letting software act on its own, booking appointments, chasing leads, or approving a payment, a decision made in Beijing this week now sits ahead of anything Putrajaya has on the books. On 15 July, China became the first country to write a dedicated rulebook for AI agents, the autonomous systems that do not just answer questions but take actions. Malaysia is adopting the same technology quickly, yet still has no equivalent law.
Editor
Kai T chevron_right
Table of Contents
What China actually did
The rules, formally the Implementation Opinions on the Standardised Application and Innovative Development of Intelligent Agents, were issued jointly by the Cyberspace Administration of China, the National Development and Reform Commission, and the Ministry of Industry and Information Technology. They took effect on 15 July 2026 and, according to AI governance trackers covering the release, create the world's first regulatory category built specifically for AI agents.
The Opinions define an agent as a system that can perceive, remember, decide, interact, and execute on its own. From there they set a three-tier decision authorisation framework, so an agent's freedom to act is tied to how much risk its actions carry. Agents used in sensitive sectors such as healthcare, transport, media, and public safety face mandatory filing, compliance testing, and even product-recall provisions if they go wrong. On the same day, China's separate rules on humanlike AI companions also took force, pushing firms including ByteDance and Alibaba to switch off some anthropomorphic features.
Why this lands in Malaysia
Malaysian companies are not watching agentic AI from a distance. Mastercard switched on AI-agent payments in Malaysia in June 2026, letting authenticated agents buy on a consumer's behalf, which means an agent here can already move money. Surveys through 2026 show the appetite: about 86% of Malaysian business leaders say they are confident about using AI agents to expand workforce capacity within 18 months, and roughly a third plan to adopt AI within two years, among the highest rates in the Asia Pacific.
The readiness underneath that appetite is thinner. Studies this year put the share of Malaysian organisations fully prepared to deploy AI at around 13%, and only about 19% have fully centralised their data, the single biggest technical block to letting an agent act across departments. Manufacturing, banking, and the public sector are furthest along. Malaysian-built agents also carry a real local edge, native support for Bahasa Malaysia, English, Mandarin, and Tamil in one system.
The gap Malaysia has not closed
Here is the uncomfortable part. Malaysia's own AI Governance Bill, drafted by the National AI Office on a risk-based model, is still working its way toward Cabinet, and it governs AI broadly rather than agents specifically (we covered its intellectual-property twist here). None of it yet answers the questions China just tried to: who is accountable when an autonomous agent makes a bad call, how much authority an agent should hold before a human signs off, and which high-stakes uses need to be registered first. As agentic commerce goes live locally, those questions stop being academic.
Malaysia does not need to copy China's rulebook, and plenty of it reflects Beijing's own priorities around control. But the direction of travel is now set. Deploying agents into banks, factories, and government offices without deciding who answers for their actions is a bet. The technology is already acting. The rules are what is running late.
Images courtesy of Igor Omilaev and Homa Appliances on Unsplash.