Artificial Intelligence Will Not Save the SDGs on Its Own — Policy Has to Catch Up First

With less than four years remaining until the 2030 deadline, the Sustainable Development Goals (SDGs) are significantly off track. In response, artificial intelligence (AI) has increasingly been presented as a means to accelerate progress. AI can forecast flood risks from incomplete hydrological data, model the spread of infectious diseases, optimize agricultural production, predict commodity prices to support the energy transition and extend legal and financial services to underserved communities. These capabilities are genuine and increasingly well documented.

Yet there is a fundamental misunderstanding at the heart of much of the discussion about “AI for the SDGs.” The challenge is not whether AI can deliver technical solutions. It is whether governance systems can direct those solutions toward socially desirable outcomes. AI is a tool. It is not a policy.

This distinction is critical. Every successful AI application depends on a series of governance choices about rights, responsibilities, accountability, and distribution. Technology can generate predictions, recommendations and decisions, but it cannot determine who benefits, who bears the risks, or who is accountable when things go wrong.

Capability does not equal governance

Consider water management, one of the most promising areas for AI deployment. Machine learning systems now outperform many traditional forecasting methods in predicting water demand and managing scarce resources. These systems can help utilities improve efficiency, reduce waste, and strengthen resilience to drought.

Efficiency and fairness are not the same, and policy frameworks too often address only the former.

However, the same predictive infrastructure can also be used to ration access in ways that reinforce existing inequalities. An algorithm can identify who should receive less water during scarcity, but it cannot determine whether that outcome is socially acceptable. Those judgments require rules established through governance.

Questions such as who sets allocation thresholds, who audits outcomes, and who has the right to challenge decisions lie entirely outside the model. Without answers to these questions, improved predictive accuracy may coexist with growing social injustice. Efficiency and fairness are not the same, and policy frameworks too often address only the former.

When explainability is not enough

The governance challenge becomes even more apparent when AI intersects with legal systems and public institutions. Predictive policing systems, risk-scoring tools, and algorithmic assessments are already used in bail, sentencing, parole and law enforcement decisions across various jurisdictions. These applications directly affect SDG 16, which seeks to promote justice, accountability, and strong institutions.

Considerable attention has been devoted to making these systems explainable. Modern AI techniques increasingly enable users to identify which variables influenced a particular decision or score. Explainability is undoubtedly valuable, but it is not equivalent to fairness, accuracy, or legitimacy.

Transparency and accountability are related concepts, but they are not interchangeable. Confusing one for the other is among the most common governance failures in contemporary AI policy.

An algorithm can clearly explain why it reached a conclusion, yet still produces a discriminatory or unjust outcome because its reasoning is based on biased or incomplete historical data. The right to due process is not satisfied merely because a machine can show its calculations. It is satisfied only when individuals have meaningful opportunities to challenge decisions and when independent institutions verify that the system operates within acceptable legal and ethical boundaries.

Transparency and accountability are related concepts, but they are not interchangeable. Confusing one for the other is among the most common governance failures in contemporary AI policy.

Governance gaps that threaten SDG progress

First, the gap between technical innovation and legal accountability. Technology evolves far more rapidly than the law. Technical standards organizations can issue guidance for autonomous vehicles, AI-assisted healthcare systems, automated public services, and other emerging technologies long before legislatures can enact comprehensive regulations. These standards play an important role in promoting safety and interoperability.

However, standards do not establish legal responsibility. When an autonomous system causes harm, societies must grapple with difficult questions. Is responsibility borne by the developer who designed the algorithm, the company that deployed it, the hardware manufacturer, or the operator overseeing the system? Only law can provide authoritative answers.

Policymakers do not face a choice between enabling innovation and regulating it. Rather, they must recognize that technological and governance questions are inseparable. 

Second, the gap between global development and local consequences. AI development is concentrated in a relatively small number of countries and corporations, yet its social and economic consequences are felt locally. The exclusion of low-resource languages illustrates the problem. Of the roughly 7,000 languages spoken worldwide, only a small fraction is meaningfully represented in contemporary AI systems.

Third, the gap between digital progress and environmental sustainability. The environmental footprint of AI remains one of the least-discussed aspects of technology’s rapid expansion. Advances in AI depend on energy-intensive data centres, increasingly powerful computing hardware and global supply chains for semiconductors and electronic components.

Governing AI for sustainable development

Technical capability is advancing faster than the legal, institutional, and governance frameworks needed to guide it. As a result, societies increasingly can deploy powerful AI systems before they have established clear rules governing their use.

Policymakers do not face a choice between enabling innovation and regulating it. Rather, they must recognize that technological and governance questions are inseparable. Every AI system deployed to support sustainable development raises questions about accountability, fairness, representation, environmental sustainability and human rights.

If AI is to accelerate progress toward the SDGs, governments must move beyond celebrating technical breakthroughs and confront the harder task of building institutions capable of guiding them. The world does not need AI alone. It needs governance systems that ensure AI serves the public good.

This article is based on a keynote address delivered at the City College of New York in June 2026.