The Three Temptations Facing the UN's First Global AI Dialogue in Geneva

On July 6–7, governments, technology companies, researchers, civil society organizations, and international institutions will gather in Geneva for the first United Nations Global Dialogue on Artificial Intelligence. The agenda will focus on familiar questions of AI governance, safety, innovation, and international cooperation. But an even more important question has received far less attention: what kind of geopolitical moment is this Dialogue taking place in, and how should that reality shape the conversation?

That question matters more than the agenda itself.

The Global Dialogue arrives at a paradoxical moment. Never has international cooperation on AI been more urgent; yet rarely have the geopolitical conditions for cooperation been less favorable. AI has become the defining technology of our age at precisely the moment when the international system is moving away from collaboration and towards strategic competition, industrial policy and transactional diplomacy. This tension, and not AI itself, should dominate the discussions in Geneva.

History offers a useful lesson. Geneva has often served as the place where dialogue began before trust existed. When Ronald Reagan and Mikhail Gorbachev met there for the first time in 1985, the Cold War was entering one of its most dangerous phases. Their summit produced no grand agreement and resolved few immediate disputes. Its significance lay elsewhere. It demonstrated that dialogue need not wait for geopolitical rivalry to subside. On the contrary, sustained dialogue became necessary because rivalry had become too dangerous to manage in silence.

AI presents a similar challenge. Unlike previous waves of technological innovation, AI is no longer discussed primarily in terms of productivity or economic growth; it has become strategic infrastructure. Governments increasingly view frontier models, advanced semiconductors, compute capacity, cloud infrastructure and data ecosystems as instruments of national power. AI policy is now inseparable from national security, industrial competitiveness and geopolitical influence.

This shift is understandable. No responsible government can ignore the implications AI will have for economic resilience, military capability, cybersecurity or scientific leadership. The problem is not that AI has become strategic. The problem is that the strategy of individual nations increasingly appears to be foreclosing the possibility of global governance.

This matters because the pace of AI development continues to accelerate while the institutions capable of governing its risks struggle to keep up. The imperative for cooperation has therefore become stronger precisely as the incentives for cooperation have weakened. Governments are investing in sovereign AI capabilities while technology partnerships are increasingly shaped by geopolitical alignment. Export controls, investment restrictions and competition over critical supply chains have become normal instruments of statecraft.

None of these developments are temporary. They reflect a structural shift in international politics. The challenge for Geneva is therefore not to reverse this trend because it cannot. Rather, it is to demonstrate that meaningful cooperation remains possible within it. That requires resisting three temptations.

The first is to avoid talking about security.

Security has become politically uncomfortable in international AI discussions because it immediately exposes competing national interests. Yet pretending security does not belong in AI governance would render the Dialogue largely irrelevant. Questions surrounding military applications, cyber resilience, frontier model security, critical infrastructure and supply chains are central to how governments understand AI today. Ignoring them will not make them disappear; it will simply ensure that the most consequential conversations take place elsewhere. A stark reminder of this reality is the recent US decision regarding Anthropic, where an export control directive forced the company to pull its powerful Mythos and Fable models offline worldwide over government cybersecurity fears.

The second temptation is the opposite: allowing security to become the entire conversation.

This would be an equally serious mistake. Once every governance question becomes a national security question, cooperation inevitably narrows, transparency becomes more difficult and technical collaboration becomes more constrained. Governance risks becoming little more than negotiation between competing strategic blocs. That would undermine one of the Dialogue's most important ambitions: ensuring that different governance approaches remain interoperable even where they are not identical.

Interoperability is often described as a technical objective, but it is rapidly becoming a geopolitical one. Countries will not converge around a single model of AI regulation, nor should they. Different legal systems and political traditions will inevitably produce different approaches. The challenge is ensuring that these systems remain sufficiently compatible to support scientific collaboration, trade, innovation and cross-border trust. Fragmentation may appear strategically advantageous in the short term. In the long term, it risks creating competing technological spheres with diminishing capacity to cooperate on shared risks.

The same logic applies to human rights. Too often, human rights are treated as the normative chapter of AI governance rather than one of its strategic foundations. Yet societies will ultimately judge AI not only by what it can do, but by whether it remains accountable to people. Transparency, due process, non-discrimination and meaningful human oversight are not obstacles to innovation; they are essential to the legitimacy that durable governance requires.

The third temptation, and perhaps the most consequential, is to treat the Global South primarily as the object of geopolitical competition.

Much of today's AI diplomacy is framed through the lens of strategic rivalry, particularly between the United States and China. As a result, developing countries are too often viewed as constituencies to be persuaded or partners to be recruited into competing technological ecosystems.

This fundamentally misunderstands the politics of AI. Countries across Africa, Latin America, Southeast Asia and other developing regions are not merely choosing between competing governance models. Many are asking different questions altogether. How can AI support economic development? Where will compute infrastructure come from? How can local languages be represented in foundation models? How can countries build the skills, institutions and regulatory capacity necessary to participate meaningfully in the AI economy?

These are not secondary concerns. They are central to whether AI governance will ultimately be viewed as legitimate and inclusive.

China has understood this dynamic, pursuing partnerships centered on infrastructure, investment and capacity-building across much of the developing world. Whether these initiatives ultimately strengthen global cooperation or reinforce geopolitical fragmentation remains an open question. What is clear, however, is that influence today is built as much through practical engagement as through diplomatic rhetoric.

Other actors should take note. The Global South should arrive in Geneva not as diplomatic terrain over which influence is exercised, but as an equal participant whose priorities help shape the conversation itself. Listening may prove more consequential than persuading.

No one should expect the first Global Dialogue to produce consensus on AI governance. That was never its purpose. Its success should instead be measured by something both more modest and more significant: whether governments remain willing to engage one another on a technology that is increasingly viewed through the lens of strategic competition. Geneva has demonstrated before that dialogue does not require trust to begin. It requires recognition that the costs of refusing to talk have become greater than the risks of engagement. That may be the most important lesson for the AI debate.

In a few days, the world's AI governance community will gather in Geneva. They should arrive prepared to discuss algorithms, models and regulation; but they should leave having answered a more fundamental question. Can the international system still cooperate on transformative technologies before geopolitical competition hardens into permanent fragmentation?

The answer will not determine the future of AI. It may, however, determine whether AI becomes another force accelerating geopolitical division or one that reminds us why international cooperation remains indispensable.