System responsibility remains human: Ti Wei Ang explains UNESCO’s approach to AI governance without stifling innovation

UNESCO says the clock has not yet run out for governments to impose effective governance frameworks on AI technologies, rejecting the hypothesis that legislative control will be lost amid accelerating technical progress. Ti Wei Ang, the organization’s specialist for ethics programmes in biology, science and technology, explained in the Frankly Speaking programme that the rapid pace of innovation does not mean slowing regulatory action, but calls for moving from reducing oversight to merely rigid legislative texts toward building a comprehensive system of standards, institutional capacities, continuous assessments of system impact, and joint international cooperation.

This view coincides with the growing internal warnings in major global development labs about the risks of smart model runaway.The tech sector recently saw the resignation of researcher Jacob Coxon, who previously worked on pre-training research at OpenAI and Anthropic, accusing both firms of a uncalculated rush toward developing superintelligence capable of self-improvement, system penetration, reshaping industry structures and gaining influence in the real world. While Ang referred to undisclosed convictions among sector workers that catastrophic risks could arise before the end of the decade, he affirmed that addressing these concerns does not involve halting technological progress, but creating a responsible environment that balances protection from severe risks with avoiding over-regulation that would push investment toward other markets.

The international organization defends the non-binding nature of its ethical recommendations, arguing that the flexibility of this framework has allowed states to keep pace with the technological explosion without being constrained by rules that could become outdated quickly, while the prospect of a binding international treaty remains subject to member-state decisions and the maturity of their dialogues. With advanced computing power and data concentrated in a limited number of countries and companies, Ang emphasized the need to prevent the private sector from monopolizing the governance compass, noting that the smart systems themselves bear no legal or ethical responsibility. According to UNESCO’s view, AI exists to support human decision-making rather than to eliminate accountability, meaning that full accountability rests solely on the individuals and organisations that design, train, deploy and monitor the systems.

Building a societal trust environment represents the real guarantee for sustainable innovation and its broad adoption.The Saudi experience stands out as a model that combines investment in technical capabilities with the establishment of institutional infrastructure through the Saudi Data and AI Authority (SDAIA) and the creation of the International Centre for AI Research and Ethics (IAC) under UNESCO’s umbrella. The Kingdom is among the first countries to have applied a readiness-assessment methodology to verify the conditions for responsible publishing, culminating in Riyadh hosting the Global Forum on AI Ethics from September 14 to 17, an event that Ang says should be judged not by the volume of data released but by the resulting implementation partnerships and deeper understanding of the practical challenges each country faces.

For companies and tech entities operating in the Gulf, Egypt and the Arab Levant, this approach shifts the notion of compliance from theoretical anticipation to daily, direct regulatory practice. Institutions in critical sectors such as health care, financial services and public services will not be able to blame algorithms if biased decisions or operational errors occur, making the establishment of human-audit pathways and impact-assessment procedures an essential prerequisite before deploying any model into a live environment. Moreover, the absence of a strict international treaty and the reliance on flexible frameworks give regional regulators an incentive to develop local guidance, requiring engineering team leaders to acquire skills in model documentation and data tracing to ensure alignment with readiness tests without delaying launch schedules.

The real test for policymakers remains converting the recommendations into operational standards that protect the public interest.AI is not inherently good or evil; its outcomes are determined by the human choices that govern its design and use, and by institutions’ ability to track its consequences at a pace that ensures safety without seizing development opportunities.