(July 23): Malaysia's proposed AI Governance Bill arrives at a decisive moment. Artificial intelligence (AI) is moving into credit scoring, hiring, insurance, education, healthcare, logistics and public administration. And its environmental footprint is growing. For business, this looks like productivity. For society, it changes who decides, who explains, who benefits and who pays the hidden cost.

The ongoing public consultation offers a chance to shape these rules before habits harden into infrastructure. In many ways the proposed Bill points in the right direction: central coordination, sectoral leadership, risk-based obligations, incident reporting and sandboxes. Its principles on human dignity, transparency, accountability, safety and data stewardship matter. Yet three additions would make the framework stronger: double literacy, so that people understand both AI and themselves well enough to retain real agency, the ProSocial AI Index, which can measure whether AI serves people, and not just efficiency, and explicit attention to AI's environmental footprint. These points are important for several reasons:

AI systems carry assumptions, targets and incentives. A hiring tool may optimise for speed and reproduce old exclusion. A chatbot may shorten waiting times yet confuse elderly customers. A credit model may reduce defaults but misread or completely miss informal income. In each case, the tool improves a metric. The question is whether it serves people. 

This is why governance cannot sit only in a regulator's office or a compliance department. It must also live in the mind of the person who buys, configures, supervises or trusts the system. The Bill defines duties for developers and deployers, requires risk assessment and incident reporting. Yet no statute can think on behalf of a citizen, teacher, doctor, banker, entrepreneur or civil servant at the moment when an AI recommendation feels convenient and authoritative.

Human agency is not protected by declaring that a human remains “in the loop”. A tired employee who clicks “approve” on an algorithmic recommendation without understanding it is not exercising meaningful oversight. A consumer who accepts a personalised offer without noticing the behavioural design around it is not freely choosing in any serious sense.

This is the purpose of double literacy. Human literacy is the ability to know one’s own mind: what I value, what I am tempted to outsource and where my attention and judgement need protecting. Algorithmic literacy is the ability to read the system: what it optimises for, what data shaped it, whose interests it serves and where its blind spots lie. 

When taken together, these literacies can make human oversight real. They belong in schools, universities, civil-service training, board education, staff training and procurement teams. Malaysia should not build AI governance that assumes agency will survive on its own. The Bill should make double literacy an explicit national capability, tied to education, sectoral guidance and capacity-building by the future AI authority.

Principles are necessary. They are not enough. Every organisation can claim to be responsible, trustworthy and human-centred. The business question is simpler: can it show the evidence? Environmental, social and governance (ESG) reporting has led to a wave of greenwashing, but it began the shift from measuring efficiency alone to measuring wider impact. AI governance can learn from both its promise and its failures.

Malaysia needs a common measurement tool that moves AI governance from aspiration to operating discipline. The ProSocial AI Index offers one practical model. It assesses both the how and the why of AI impact. The how asks whether the system is tailored to its real context, trained on representative, fit-for-purpose data, tested against real conditions, and targeted to a clearly stated outcome. The why asks where the value goes across four domains: purpose, people, prosperity and planet.

Together these form a sixteen-cell grid that procurement officers, boards, regulators and auditors can use before, during and after deployment. A bank using AI for credit scoring would need to show not only accuracy, but also fairness for informal workers and small and medium enterprises. A university using AI in assessment would need to show that it promotes learning with build-in friction, not merely fraud detection. A public agency would need traceability, clear routes of appeal and human review.

The advantage of an index is comparability. Without a shared yardstick, every ministry, company and vendor invents its own definition of responsible AI. With a common dashboard, Malaysia can map, measure and manage AI systems across sectors, while giving innovators a clearer path to trust, and giving the country a compass to track whether the vision of an AI Nation actually serves people and planet. 

The AI governance conversation rightly focuses on people, rights, accountability and safety. It must also include the planet. AI has a physical body. It runs on electricity, water, minerals, land, chips, cooling systems and data centres. As Malaysia courts digital investment, the question is what kind of growth the country will reward.

An AI model that reduces emissions in logistics, improves grid management or supports precision agriculture can create real long-term value. A compute-heavy system that produces marginal convenience while increasing energy and water demand will not. Environmental impact should be part of risk assessment, procurement, sandbox evaluation and reporting for significant AI systems. Resource use is not an externality; it is part of the design brief.

Malaysia can lead by asking a better question than “Is this AI safe enough to deploy?” The stronger question is: “What does this AI help restore, strengthen or regenerate?” Does it increase human capability and widen fair access? Does it reduce ecological pressure and preserve local eco-environments or simply line the pockets of the tech billionaires for relatively limited returns for the nation and its people?

This is what regenerative intent means; building these questions in from the start, into procurement criteria, vendor contracts, data governance, human oversight, energy reporting and board accountability, rather than adding them as a corporate social responsibility paragraph after deployment. 

This is also a Malaysian advantage waiting to be claimed. The country has a rich moral vocabulary of amanah, maslahah, stewardship, responsibility, shared prosperity and community duty. The AI Bill can translate that vocabulary into practical governance.

Readers should use the public consultation to ask for three concrete improvements. First, embed Double Literacy in human oversight, education and institutional capacity-building, so citizens and professionals keep their judgment in an AI-mediated economy. Second, adopt a shared ProSocial AI Index to measure how AI systems perform and who they serve, across purpose, people, prosperity and planet. Third, make environmental footprint and regenerative intent explicit in AI risk policies, procurement, sandbox standards and reporting requirements. Feedback can be submitted through the Unified Public Consultation portal: https://upc.mpc.gov.my/view-consultation/264. The deadline is July 31, 2026. 

Malaysia's AI future should not be shaped only by those who build systems or buy them. It should also be shaped by those whose lives, work, water, energy and choices those systems will touch, and that means all of us. 

Dr Cornelia C Walther is associate professor at Sunway University's Institute for Global Strategy and Competitiveness and senior fellow at the Sunway Centre for Planetary Health. She advises UNFPA on hybrid intelligence and is a senior fellow at Harvard and the Wharton School.