SEPTEMBER 21 — Artificial intelligence has become a field in which terminology often races ahead of reality.
Generative AI, agentic AI, artificial general intelligence and superintelligence are increasingly spoken of as though they belong to the same technological continuum and carry essentially the same dangers. They do not.
Generative AI can produce text, images, computer code, audio and increasingly sophisticated combinations of them.
Agentic AI goes further by allowing systems to pursue specified objectives through sequences of actions, sometimes with limited human intervention.
Both create serious problems.
Hallucination, cyber vulnerability, manipulation, privacy breaches, labour displacement, concentration of computational power and excessive human dependence on automated decision-making deserve sustained attention.
But these are not necessarily systemic risks in the strongest meaning of the term.
The deeper threshold would be crossed with what can loosely be called Super AI: an artificial intelligence possessing capabilities vastly beyond those of human beings across numerous intellectual domains, potentially combined with a capacity for recursive self-improvement.
The distinction is fundamental. Generative AI generates. Agentic AI acts within assigned parameters. Super AI, if it becomes technically possible, could potentially improve the very architecture through which it thinks and acts. That changes the nature of the problem.
The central issue is no longer simply whether an AI system produces an incorrect answer or whether an autonomous agent performs the wrong task.
It becomes whether human institutions can retain meaningful control over systems capable of improving their capabilities at speeds that regulation, governments and perhaps even their creators cannot match.
This is where AI risk becomes systemic.
A financial crisis becomes systemic when difficulties originating in one institution spread throughout the financial architecture.
A pandemic becomes systemic when contagion overwhelms the ability of individual jurisdictions to contain it.
Super AI could become systemic for similar reasons: its consequences may not remain confined to the laboratory, company or country where it originated.
This possibility becomes especially serious when combined with geopolitical competition.
The United States, China, Europe and other technological powers increasingly regard advanced AI not merely as commercial technology but as part of national power.
Semiconductor capacity, cloud computing, data centres, algorithms, electricity supplies and highly specialised human talent are becoming components of AI sovereignty.
Asean cannot remain outside this discussion.
Singapore and Malaysia are developing important positions within the regional AI and semiconductor ecosystems.
Indonesia, Vietnam, Thailand and the Philippines are also building digital capabilities. Asean therefore sits increasingly close to the infrastructure through which the global AI economy operates.
Yet Asean should resist treating every form of artificial intelligence as an existential danger.
Over-regulating generative and agentic AI because of hypothetical dangers associated with superintelligence could impose substantial economic costs.
Southeast Asia still needs AI to increase productivity, improve healthcare, modernise education, strengthen logistics, expand financial inclusion and help smaller companies compete internationally.
The correct response is therefore differentiation rather than panic.
Generative AI requires standards concerning accuracy, copyright, privacy, transparency and responsible deployment.
Agentic AI requires stronger safeguards because autonomous actions can produce consequences beyond the generation of information. Human oversight, audit trails, permission boundaries and emergency shutdown mechanisms become increasingly important.
Super AI would require another level of governance altogether.
The danger would become especially acute if recursive self-improvement allowed an advanced system to redesign portions of itself, test those improvements and repeat the process.
Whether such a technological trajectory is achievable remains uncertain.
But precisely because its consequences could be extraordinary, governments should begin discussing the governance architecture before the technology reaches that point.
The analogy with nuclear weapons is imperfect but instructive.
Humanity did not abolish nuclear physics because nuclear weapons were dangerous.
It differentiated peaceful nuclear applications from weapons systems while constructing elaborate regimes of safeguards, verification, deterrence and arms control. AI governance requires comparable conceptual discipline.
Asean should consequently support international discussions on advanced AI safety while preserving sufficient regulatory space for beneficial AI development. Its traditional preference for dialogue among competing powers could become useful here.
Washington and Beijing may disagree profoundly over technology, semiconductors and strategic primacy, yet neither has an interest in losing control of increasingly autonomous artificial intelligence.
This is where Asean’s circumscribed power may nevertheless prove consequential.
The region cannot determine the trajectory of global AI by itself.
But it can advocate clearer distinctions between ordinary machine learning, generative AI, agentic AI, artificial general intelligence and potential superintelligence.
Without these distinctions, public policy risks oscillating between complacency and alarmism.
The most immediate task is not to fear every chatbot or autonomous software agent. It is to understand where the technological thresholds lie.
Generative AI is already transforming how human beings create knowledge. Agentic AI may transform how organisations execute decisions. Both require serious governance.
Super AI represents something categorically different.
If machines eventually acquire the ability not merely to assist human intelligence but to surpass it broadly, improve themselves recursively and operate across interconnected digital systems, humanity would confront a systemic risk of an altogether different order.
The challenge is therefore neither to stop AI nor to surrender to it.
It is to ensure that as intelligence becomes increasingly artificial, authority remains unmistakably human.
* Phar Kim Beng is a professor of Asean Studies at International Islamic University Malaysia (IIUM) and a director at the Institute of International and Asean Studies (IINTAS).
** This is the personal opinion of the writer or publication and does not necessarily represent the views of Malay Mail.
