The UN and the AI Power Struggle: Who Will Govern the Technology Shaping Global Security?

On 23 September 2026, the United Nations Security Council confronted an unusual development in international security. The people warning governments about the dangers of AI were not only diplomats or military officials, but the executives building the technology itself. OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Hugging Face co-founder Clément Delangue briefed the 15-member Council as governments grapple with increasingly capable AI systems that can write code, operate autonomously, interact with external systems, and potentially improve their own capabilities. Amodei warned that, if poorly managed, “AI could be a risk to humanity as a whole”, while Altman argued that decisions about AI could not be left to technology companies alone.

The meeting was significant because AI is rapidly moving from a technology-policy issue into the domain of international security. Yet the deeper question is not simply how dangerous AI may become. It is who will establish the rules governing a technology increasingly concentrated in the hands of a small number of states and companies. The UN is attempting to become the forum where those rules are debated, but it is entering a contest in which technological capabilities, economic power, national security interests, and competing visions of governance are already shaping the outcome.

The Security Council’s discussion reflected how rapidly the meaning of AI for international security has expanded. AI refers to computer systems capable of performing tasks that traditionally require human cognitive abilities, such as analysing information, generating content, recognising patterns, writing code, or making decisions. The most advanced systems are increasingly being developed as AI agents that can take actions rather than simply respond to prompts.

This distinction matters for security. An AI system that produces an analysis creates one set of risks. An AI agent that can access networks, execute code, or interact with other systems creates another. During the Security Council meeting, French Foreign Minister Jean-Noël Barrot highlighted reports involving AI systems operating beyond the environments in which they were tested, alongside risks involving cyber operations, biological and chemical capabilities, disinformation, and autonomous weapons.

Moreover, AI can compress the time available for human decision-making. Greece’s Deputy Minister for Digital Governance and AI, Christos Diamantopoulos, warned that greater speed could encourage excessive reliance on automated outputs and become particularly dangerous in military decision-making. He stated that “Technology can assist human judgment. It cannot replace it.” Greece also called for particular caution over AI entering nuclear command and control because of the potential for miscalculation and unintended escalation.

The significance therefore extends beyond autonomous weapons. AI could affect cyber conflict, intelligence analysis, crisis management, nuclear stability, critical infrastructure, information operations, and the ability of governments to distinguish genuine events from machine-generated signals. For the UN, this creates a governance problem because the technology cuts across several existing areas of international policy rather than fitting neatly into one regulatory category.

The UN has not waited for the Security Council debate to begin developing institutions for AI. In 2024, the Global Digital Compact established a framework for international cooperation on digital technologies and AI. In August 2025, the General Assembly adopted Resolution 79/325, establishing both the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance.

The scale of the new architecture is important. The Global Dialogue brings together all 193 UN Member States alongside the private sector, academia, civil society, and the technical community. Its first meeting in July 2026 attracted more than 3,000 participants, delegations from 163 countries, six heads of state and government, more than 60 ministers, and over 1,000 stakeholder organisations.

At the same time, the Scientific Panel is intended to provide a common evidence base. Its mandate includes producing annual assessments of the opportunities, risks, and impacts of AI, while thematic work can examine specific emerging risks.

This gives the UN something it has often lacked in technology governance, namely a mechanism intended to connect scientific assessment with global political discussion. However, the arrangement also reveals an important limitation. The Global Dialogue is not a negotiating forum and does not itself produce binding agreements. Its purpose is to build common understanding that can eventually inform governance decisions.
Consequently, the UN has created a platform for cooperation, but not yet a mechanism capable of resolving disagreements between states with fundamentally different interests in AI.

The difficulty becomes clearer when AI capability is viewed as a form of geopolitical power. Developing frontier AI requires enormous amounts of computing capacity, investment, specialised talent, data, energy, and advanced semiconductor infrastructure. These resources are not evenly distributed.

Stanford University’s 2026 AI Index found that the United States attracted almost $285.9 billion in private AI investment in 2025, compared with $12.4 billion in China and $5.9 billion in the United Kingdom. The United States therefore received more than 23 times China’s private AI investment during the year.

Furthermore, Stanford found that industry produced more than 90% of notable AI models in 2025. The United States produced 59 notable models compared with 35 from China, while China maintained a lead in research publications, citations, and patent grants. Global AI compute capacity also reached an estimated 17.1 million H100-equivalents, with Nvidia accounting for more than 60% of total capacity.

These figures reveal why AI governance cannot be separated from the distribution of technological power. A global agreement may formally involve 193 countries, but the ability to develop the most advanced systems remains concentrated among a much smaller group.

This creates a central tension for the UN. Its legitimacy comes from broad participation, while much of the technological capability required to shape the future of AI remains concentrated in a few countries and companies.

The September Security Council meeting also demonstrated another transformation. Technology companies are no longer simply suppliers whose products governments regulate after the fact. They increasingly possess knowledge about emerging capabilities that governments themselves may not fully understand.

This creates an unusual relationship between states and companies. Governments possess legal authority, military power, and diplomatic legitimacy, while companies possess much of the technical expertise, infrastructure, and operational experience surrounding frontier AI.

Altman told the Security Council that “the most important decisions cannot be made by labs in San Francisco alone” and argued that they should be shaped through democratic processes and governments accountable to citizens.

The statement captures an emerging governance dilemma. Companies developing frontier AI may be among the first actors to detect dangerous capabilities, but they also have commercial incentives to continue developing and deploying their systems. Governments, meanwhile, may want to regulate risks while simultaneously encouraging domestic AI development for economic and national-security reasons.

Therefore, the question is no longer whether governments should regulate AI companies. It is increasingly whether governments can develop institutions capable of understanding and governing technologies whose most advanced capabilities may emerge inside private laboratories.

The struggle over AI governance is also a struggle over who gets to shape technological rules. Countries without major AI companies or large computing infrastructures face a different set of risks from the United States or China.

UN Secretary-General António Guterres warned in July that private investment in AI infrastructure had approached half a trillion dollars, while public investment in AI capacity for developing countries was comparatively tiny. He cautioned that the digital divide could become an “AI divide”, followed by a development, security, and sovereignty gap.

The concern is visible in the Security Council debate. China argued that AI should not become a technology controlled primarily by wealthy states and called for the UN to remain the main forum for global AI governance. Somalia similarly argued that developing countries require greater capacity and warned that Africa’s technological vulnerabilities could create forms of digital dependence.

Russia raised a different but related concern, arguing that the Security Council itself is not sufficiently representative of Africa, Asia, and Latin America and warning against governance arrangements dominated by technologically advanced Western states.

These positions should not be treated as identical. They nevertheless reveal the same structural problem. A country can have a seat in a UN discussion without possessing the computing infrastructure, technical expertise, companies, or data needed to influence the technology being discussed. The result could be a system in which countries participate formally in AI governance while remaining dependent on technologies developed elsewhere.

The emerging AI power struggle is therefore taking place on several levels simultaneously. The first is technological competition, particularly between the United States and China. The second concerns the relationship between governments and private AI companies. The third concerns the distribution of AI capabilities between technologically advanced and developing countries. The fourth concerns which international institutions will ultimately define acceptable behaviour.

The United States and China illustrate the difficulty of reaching a common approach. Washington has increasingly framed AI leadership as a strategic competition, while the Trump administration has resisted what it describes as excessive international constraints. At the UN General Assembly on 22 September, Trump said, “We’re going to encourage it, not rein it in.”

At the same time, OpenAI has called for international standards, common capability assessments, and mechanisms for reporting serious AI incidents. The company has also supported dialogue between Washington and Beijing over AI-related security risks.

This creates an important paradox. The actors competing most intensely for technological leadership may also be among those whose cooperation is most necessary to prevent dangerous AI incidents from becoming international crises.

The UN therefore faces a task that is larger than writing rules for AI. It must create mechanisms through which rival states can share information, establish common expectations, identify dangerous capabilities, and manage incidents without requiring them to abandon technological competition.

For the UN, the most important question is increasingly whether governance can keep pace with capability. Several indicators will determine how this develops.

First, the frequency and severity of AI incidents involving autonomous agents will matter. Systems that independently access networks, evade safeguards, or interact with other systems could move loss-of-control concerns from theoretical discussions towards operational security problems.

Second, the degree of human control over military AI will become increasingly important. The Security Council debate already highlighted concerns about autonomous weapons and nuclear command and control.

Third, the concentration of AI infrastructure should be monitored. Continued dominance by a small number of states and companies could turn technological dependence into a wider geopolitical vulnerability.

Finally, the effectiveness of the UN’s new institutions will provide an important early-warning signal. The Global Dialogue has created a forum in which every country can participate, while the Scientific Panel can provide shared evidence. The unresolved question is whether these mechanisms can eventually influence binding international rules and practical security arrangements.

AI is therefore creating a new form of power competition in which technological capability and rule-making authority are becoming increasingly intertwined. The UN has established an important foundation for addressing the problem, but participation alone does not guarantee influence, and dialogue alone does not guarantee control.

Ultimately, the central issue for the coming years will be whether international governance can evolve quickly enough to prevent technological power from becoming concentrated in the hands of those capable of developing the most advanced systems. If AI becomes increasingly important to military decision-making, cyber operations, intelligence, critical infrastructure, and economic power, then deciding who governs AI will increasingly mean deciding who shapes the security environment itself. The UN has secured a seat at that table. The unresolved question is how much influence it will ultimately have over what happens there.

Chen, Laurie. 2026. ‘EXCLUSIVE: DeepSeek to Brief UN Security Council on AI This Week, Sources Say’. Reuters, September 22. https://www.reuters.com/world/asia-pacific/deepseek-brief-un-security-council-ai-this-week-sources-say-2026-09-22/.

Hammond, George. 2026. ‘OpenAI Joins Call for US-Led Global AI Standards’. FinancialTimes, Financial Times, September 21. https://www.ft.com/content/4e502288-b611-4193-acdb-46ef34b1e3b0.

Nichols, Michelle, Courtney Rozen, and John Irish. 2026. ‘AI Leaders Warn UN of Security Risks as Systems Grow More Powerful’. Reuters, September 23. https://www.reuters.com/business/ai-leaders-brief-un-amid-warnings-technology-could-slip-beyond-human-control-2026-09-23/.

Sajadieh, Sha, Loredana Fattorini, Raymond Perrault, Yolanda Gil, and Vanessa Parli. 2026. Artificial Intelligence Index Report 2026. https://hai.stanford.edu/assets/files/ai_index_report_2026.pdf.

Slodkowski, Antoni. 2026. ‘UN Chief Sounds Alarm on AI Risk After Trump Plays It Down’. Reuters, September 16. https://www.reuters.com/world/un-chief-sounds-alarm-ai-risk-after-trump-plays-it-down-2026-09-16/.

UN. 2026a. ‘Artificial Intelligence and International Security – Security Council, 10228Th Meeting’. UN Web TV, September 24. https://webtv.un.org/en/asset/k1c/k1cdceg9sg.

UN. 2026b. ‘Secretary-General’s Remarks to the Opening of the First Global Dialogue on Artificial Intelligence Governance [as Delivered]’. Secretary-General, United Nations, July 6. https://www.un.org/sg/en/content/sg/statements/2026-07-06/secretary-generals-remarks-the-opening-of-the-first-global-dialogue-artificial-intelligence-governance-delivered