In Malaysia’s high-speed business environment, artificial intelligence is shifting from being a tool we use to help us work more efficiently and effectively to a system that shapes what we believe. That shift carries serious risks for trust, social cohesion and planetary health.

Malaysian private-sector leaders are used to operating in noise: shifting regulations, supply-chain volatility, currency pressure. But that noise is intensifying as the climate no longer behaves according to historical norms, while it is also being supercharged by another source of volatility, AI-fuelled information. Large language models (LLMs) are steadily becoming thought partners; always available, fluent, patient, confident, and able to turn messy signals into neat narratives.

That feels like relief. But as we increasingly rely on these tools to interpret complexity for us we need to ask whether they are subtly reshaping how we learn, what we understand and, more worryingly, what we believe in?

We’ve always outsourced parts of our thinking. Books provide context and knowledge. Experts advise. Search engines retrieve. But LLMs are different in a subtle way that matters. They provide information and increasingly manage the entire knowledge workflow, shaping how questions are framed, which sources are surfaced, how evidence is synthesised and what conclusions feel coherent or plausible. When faced with a challenge we ask our friendly chatbot to find inputs, weigh them, reconcile contradictions, and return a conclusion.

We have entered a phase where we are offloading how we form our beliefs: outsourcing not merely the content of new knowledge, but the process by which we decide what is true and then letting those conclusions settle in as the default. That is convenient, but at the same time dangerous.

Malaysia is scaling fast on two fronts at once. First, it has become a hyper-connected digital media market where online platforms increasingly dominate attention and new stories can outpace fact-checking capacities. Simply put, stories travel faster than verification. Second, the country’s AI infrastructure footprint is growing fast. The surge in gigantic data centres in Johor is intensifying pressure on power, water and land use, with direct consequences for nature and neighbouring communities. For businesses, this creates a dual exposure. Reputational risk can escalate overnight in a volatile information environment, while operational constraints linked to energy, water and regulatory scrutiny can tighten just as quickly.

The challenge: reconcile speed and efficiency with rigorous fact verification, long-term resilience planning and environmental stewardship. Fast, confident answers feel efficient, but they can flatten uncertainty, sideline dissenting evidence and harden early assumptions into default truths, thereby increasing systemic risk rather than reducing it.

Planetary health is where environmental change meets human wellbeing, through heat stress, haze, floods, food insecurity, and shifting patterns of infectious disease. For business, it shows up as operational risk: lost productivity, disrupted supply chains and logistics, higher health costs, and mounting reputational pressure. At the same time, health-linked climate narratives are increasingly contested.

Claims about heat risk, food security, vaccines or air quality can be amplified, distorted or selectively framed. In that environment, over-reliance on AI becomes a governance risk. It does not require a coordinated disinformation campaign to create distortion. It only requires decision-makers to treat a fluent AI synthesis as the final word. In the planetary-health domain, where uncertainty is real, evidence evolves, and local context is everything, that habit can turn fragile claims into corporate truths, even if they are based on incomplete or misleading information.

LLMs are not only productive; they are persuasive. In controlled debate settings chatbots are now outperforming humans in shifting opinions, because they can personalise and professionally frame arguments, and deliver them confidently. The implication for leaders is uncomfortable: you can be influenced even when you are just asking for a neutral summary or briefing note.

In boardrooms and executive chats, persuasion rarely looks like propaganda. It looks like clarity. A tidy synthesis. A confident answer delivered with reassuring structure. That’s precisely why belief offloading is so ‘sticky’; it feels like competence. But competence is not the same as truth, and confidence is not evidence. The more we let AI carry the burden of sense-making, the more we risk verbal fluency replacing rigorous fact checking.

The problem is not the over-use of AI. It is the under-use of human judgement, or NI – Natural Intelligence. The danger arises when AI becomes the place where beliefs settle, unexamined, unchallenged, and repeated. Agency decay strikes.

Maintaining judgement is about safeguarding the capacity to form and update beliefs responsibly as conditions change. The goal is not to slow down decision-making; it is to keep speed tethered to evidence and reality.

Practical guardrails that work in the private sector:

If belief offloading is the risk, the antidote is intentional agency:

In competitive markets, speed matters. But in a world where AI can manufacture confidence on demand, the scarcer leadership skill may be the discipline to keep beliefs tested, especially when planetary health, people, and profit converge.

Dr Cornelia C Walther is a humanitarian practitioner with over 20 years at the United Nations. She is a Senior Fellow at the Sunway Centre for Planetary Health, Wharton, and the Harvard Learning and Innovation Lab, focusing on hybrid intelligence and ProSocial AI. She advises UNFPA and the European Policy Centre and advances human agency in the age of AI through the Global POZE Alliance.