A 20-Year Pattern of Opposition to Multilateral Technology Governance

ISAR Global Intelligence Brief
Classification: Governance Process Intelligence
Date: March 2026
Author: ISAR Global Research Directorate

Executive Summary

On 12 February 2026, the United Nations General Assembly appointed 40 members to the Independent International Scientific Panel on Artificial Intelligence by a recorded vote of 117 in favour to 2 against (United States and Paraguay), with 2 abstentions.[^1] The United States representative explicitly stated: “With sovereignty in mind, the United States wishes to register its strong objection to the establishment of the Panel as currently constituted.”^2

This opposition was not isolated. In September 2025, Michael Kratsios, Director of the White House Office of Science and Technology Policy, categorically rejected UN AI governance efforts at a Security Council debate: “We totally reject all efforts by international bodies to assert centralized control and global governance of AI.”[^3]

Core Finding: The US rejection of UN AI governance represents consistent policy spanning two decades, not an anomalous position specific to the current administration. From 1996 to 2026, the United States has systematically opposed intergovernmental control of technology infrastructure, preferring industry-led multi-stakeholder processes over UN-based frameworks.

Strategic Implication: The UN AI Panel faces a fundamental authority problem: it possesses democratic legitimacy (117-2 vote) but lacks practical influence over the country hosting the majority of leading AI companies. Historical precedent from internet governance suggests the Panel will achieve limited implementation success without US cooperation, regardless of broad international support.

The February 2026 Vote: What Happened

The Appointment Process

Following an open call for candidates that closed 31 October 2025, the UN Secretary-General received over 2,600 applications.[^4] After independent review by the International Telecommunication Union (ITU), the UN Office for Digital and Emerging Technologies, and UNESCO, Secretary-General António Guterres recommended 40 members to the General Assembly on 4 February 2026.^5

On 12 February 2026, the United States requested a recorded vote on the appointments—unusual for UN scientific panels, which typically proceed by consensus.[^6] The vote breakdown revealed stark geopolitical divisions:

The US Statement

The US representative framed opposition around sovereignty and institutional mandate:

“With sovereignty in mind, the United States wishes to register its strong objection to the establishment of the Panel as currently constituted… The Panel, as well as the associated Global Dialogue on AI Governance, are a significant overreach of the UN’s mandate and competence.”^8

The representative urged the UN to focus on “core missions—international peace and security, human rights and humanitarian assistance—rather than attempt to regulate or stifle the development of cross-cutting and cutting-edge technologies that will determine economic and strategic competition in the twenty-first century.”^9

The Developing Country Response

Uruguay’s representative, speaking for the Group of 77 and China, countered:

“We recall the bloc’s consistent call for comprehensive international frameworks that guarantee the fair inclusion of developing countries in shaping the future of AI governance… We therefore express regret that a vote has been called on such an important matter.”^10

This exchange encapsulates the core tension: developing countries seeking inclusive multilateral governance versus the United States preferring bilateral partnerships and industry-led standards.

The Kratsios Doctrine: September 2025

The February 2026 vote followed categorical rejection of UN AI governance five months earlier. On 24 September 2025, Michael Kratsios addressed the UN Security Council’s Open Debate on Artificial Intelligence and International Peace and Security, delivering the clearest articulation of US policy to date.[^11]

Core Policy Statement

“We totally reject all efforts by international bodies to assert centralized control and global governance of AI. We believe that the responsible diffusion of AI will help pave the way to a flourishing future, one of increased productivity, empowered individuals, and revolutions in scientific advancement.”^12

Governance Philosophy

“The path to this world is found not in bureaucratic management, but in the freedom and duty of citizens, the prudence and cooperation of statesmen, and the independence and sovereignty of nations.”^13

Risk Framework

“We believe broad overregulation incentivizes centralization, stifles innovation, and increases the danger that these tools will be used for tyranny and conquest. Ideological fixations on social equity, climate catastrophism, and so-called existential risk are dangers to progress and obstacles to responsibly harnessing this technology.”^14

Kratsios explicitly rejected concerns about AI existential risk as “ideological fixation,” directly contradicting positions held by many Panel members, including prominent AI researcher Yoshua Bengio, who had briefed the Security Council immediately before Kratsios spoke.[^15]

Historical Context: A 20-Year Pattern

The US rejection of UN AI governance is not novel policy. It represents the latest iteration of consistent opposition to intergovernmental control of technology infrastructure spanning two decades.

1996: ITU and the Domain Name System

In 1996, the International Telecommunication Union attempted to establish control over the Internet’s domain name system. The US administration blocked this effort, leading to the creation of the Internet Corporation for Assigned Names and Numbers (ICANN) in 1998—a private corporation under US government oversight, not an intergovernmental body.[^16]

This established the template: multi-stakeholder processes (government, private sector, civil society) rather than intergovernmental control.

2003-2005: World Summit on the Information Society

The first phase of the World Summit on the Information Society (WSIS) in Geneva (December 2003) failed to reach agreement on internet governance. Brazil, China, South Africa, and numerous Arab states advocated for greater UN/ITU control, while the US maintained that internet infrastructure should remain under existing private-sector arrangements.[^17]

The Working Group on Internet Governance (WGIG) was established to develop proposals for the second WSIS phase in Tunis (November 2005). Weeks before the Tunis summit, the US issued a statement reiterating “its historic role in authorizing changes or modifications to the authoritative root zone file.”[^18]

The US negotiating position was documented as: “flexible on the principle of global involvement, very strong on the principle of multistakeholder participation, but inflexible on the need for US control to remain for the foreseeable future in order to ensure the security and stability of the Internet.”[^19]

The compromise created the Internet Governance Forum (IGF)—a discussion platform with no decision-making authority.^20 The US successfully prevented any mechanism that would transfer control of internet infrastructure to intergovernmental bodies.

2009-2012: ITU Regulatory Attempts

Between 2009 and 2012, the ITU made repeated attempts to gain authority over internet-related infrastructure:

The WCIT-12 negotiations failed to reach consensus, with the US and European countries refusing to sign the revised treaty.^23 Russia, China, and numerous developing countries supported bringing internet governance under ITU authority; the US-led coalition successfully blocked these efforts.

The Consistent Pattern

Across 20 years and multiple US administrations (Clinton, Bush, Obama, Trump, Biden, Trump II), US policy has remained consistent:

  1. Multi-stakeholder processes (industry-led with government participation) are preferred over intergovernmental processes (UN-led with industry participation)
  2. Technical infrastructure should remain with existing private-sector arrangements, not transfer to international treaty bodies
  3. Innovation and competition arguments consistently override inclusive governance arguments
  4. Bilateral partnerships are the preferred alternative to multilateral frameworks

The 2025-2026 rejection of UN AI governance follows this established pattern precisely.

The US Alternative Model

Rather than participate in UN AI governance frameworks, the United States is constructing an alternative architecture based on bilateral partnerships and technology exports.

American AI Exports Program

Announced at the India AI Impact Summit in February 2026, the American AI Exports Program represents the US alternative to multilateral governance:[^24]

National Champions Initiative: Integration of partner nations’ leading companies into the American technology ecosystem, providing access to US AI infrastructure and capabilities.

US Tech Corps: A “Peace Corps-style” model deploying volunteer technical experts to partner countries to provide “last-mile support” for AI deployment in public services.^25

Global Financing Mechanisms: Dedicated World Bank funds, export-linked financing through the Development Finance Corporation (DFC) and Export-Import Bank (EXIM), and grant-based technical assistance to boost AI readiness in priority markets.[^26]

The “Sovereign AI” Framework

At the India summit, Kratsios articulated a conception of sovereignty that justifies US technology dependence:

“We totally reject global governance of AI… Real sovereignty means owning and using best-in-class technology—specifically the American AI stack—to benefit one’s own people while maintaining national control over data and policy.”[^27]

This frames dependence on US AI infrastructure not as constraint on sovereignty but as enablement of it—countries achieve “strategic autonomy” through leveraging American technology rather than through international frameworks that might constrain US dominance.

Bilateral Agreements

The Pax Silica Declaration, signed by the US and India at the summit, exemplifies this approach: bilateral agreements pledging cooperation on AI development “unapologetically friendly to entrepreneurship” and focused on securing the “physical silicon stack.”^28

This creates an alternative governance architecture: rather than universal frameworks negotiated at the UN, the US offers bilateral partnerships providing access to American AI infrastructure in exchange for alignment with US technology standards and policy preferences.

Geopolitical Implications

The Legitimacy Paradox

The UN AI Panel faces a fundamental contradiction:

Democratic legitimacy: The 117-2 vote demonstrates overwhelming international support, particularly from developing countries seeking inclusive governance frameworks.

Practical authority: The United States hosts the majority of leading AI companies (Google, Microsoft, OpenAI, Anthropic, Meta, NVIDIA) and controls significant AI infrastructure. US rejection limits the Panel’s ability to influence actual AI development and deployment.

This creates competing claims to legitimacy: numerical support from UN member states versus technological dominance from the US private sector.

The China Dimension

China explicitly supported the UN framework. Vice Minister of Foreign Affairs Ma Zhaoxu stated at the September 2025 Global Dialogue launch:

“It is vital to jointly foster an open, inclusive, fair and nondiscriminatory environment for technological development and firmly oppose unilateralism and protectionism. We support the U.N. playing a central role in AI governance.”[^29]

This positions China as champion of multilateralism while the US appears isolated. However, China simultaneously pursues its own AI standards and governance frameworks domestically, suggesting strategic hedging rather than full commitment to UN processes.

The Corporate Position

Early indications suggest potential divergence between US government and corporate positions. Several US-based technology companies participated in the September 2025 Global Dialogue launch despite government opposition.[^30] This raises questions about whether corporate actors will engage with Panel processes independently of government policy.

If US companies treat Panel recommendations as relevant guidance while the US government rejects them, this creates implementation pathways that historical internet governance precedents did not anticipate.

Implementation Analysis

What Can the Panel Achieve Without US Buy-In?

Historical precedent from internet governance suggests limited but non-zero impact:

Norm-setting in developing countries: The 117-country coalition supporting the Panel represents significant global territory. Panel recommendations may influence AI governance frameworks in countries lacking domestic technical capacity, creating de facto standards in portions of the Global South.

Academic and civil society influence: Panel members include prominent researchers whose work influences global AI discourse regardless of government positions. Academic networks may transmit Panel thinking independently of official channels.

Long-term legitimacy building: The IPCC (Intergovernmental Panel on Climate Change) faced initial US skepticism but eventually achieved broad acceptance. The AI Panel might follow similar trajectory, though the 20-year internet governance record suggests this is uncertain.

Limitations are severe: The Panel cannot influence US AI development practices without US cooperation. If Panel recommendations conflict with US corporate or government interests, historical precedent suggests they will be ignored.

Scenarios for Panel Effectiveness

Scenario A: Bifurcated Governance
UN framework governs AI deployment in developing countries; US bilateral framework governs AI in allied nations. This creates regulatory fragmentation rather than global coordination.

Scenario B: Corporate-Government Divergence
US companies engage with Panel recommendations despite government opposition, similar to corporate participation in Paris Climate Agreement after US withdrawal. This provides implementation pathway without government endorsement.

Scenario C: Diplomatic Theater
Panel becomes forum for expressing positions but produces limited practical impact, similar to Internet Governance Forum—inclusive discussion without decision-making authority.

Scenario D: Long-Term Legitimacy
Panel builds authority over time as AI governance challenges become more acute, eventually gaining US buy-in similar to IPCC trajectory. Historical internet governance record suggests this is least likely scenario.

Strategic Questions

For Governments

Developing countries: Does supporting UN framework despite US opposition provide meaningful governance capacity, or does it create parallel structure with limited influence over actual AI development?

US allies: How to navigate between maintaining alliance relationships and participating in inclusive governance frameworks that US opposes?

China and Russia: How to leverage US isolation to build alternative governance coalitions without appearing to promote authoritarian AI governance models?

For International Organisations

UN system: Can Panel achieve impact without major power support, or does US opposition doom it to advisory status with limited implementation?

Regional bodies: Should EU, African Union, ASEAN develop regional AI frameworks rather than rely on global UN process that US rejects?

For Corporations

US tech companies: Engage with Panel despite government opposition, or align with government position and risk excluding selves from 117-country coalition?

Non-US companies: Opportunity to influence global standards while US companies absent, or risk fragmentation that disadvantages everyone?

Conclusions

The US rejection of the UN AI Panel is neither anomalous nor surprising. It represents consistent policy opposing intergovernmental control of technology infrastructure across two decades, three technologies (internet, telecommunications, AI), and six US administrations.

Key findings:

  1. Pattern consistency: US policy has remained stable despite administration changes. Opposition to UN tech governance is institutional policy, not partisan politics.
  2. Alternative architecture: The US is not simply rejecting multilateral governance—it is building bilateral alternative through AI Exports Program and technology partnerships.
  3. Legitimacy paradox: Panel has democratic support (117-2) but limited practical authority over AI development concentrated in US companies.
  4. Historical precedent matters: 20 years of internet governance conflicts provide roadmap for AI governance trajectory. US successfully resisted multilateral control of internet; likely to replicate with AI.
  5. Corporate position unclear: Potential divergence between US government and corporate positions represents variable not present in internet governance conflicts. Could create implementation pathways if companies engage despite government opposition.

Strategic implication for decision-makers: The Panel will likely achieve influence in developing countries and academic/civil society networks but face severe limitations influencing US AI development. Organisations and governments requiring AI governance guidance should not rely solely on UN processes—bilateral engagement with US and development of regional frameworks may prove more practically effective.

This represents a case study in multilateral governance limits: votes do not equal implementation power when major technological powers opt out. Governance reality diverges from governance rhetoric.

References

[^1]: United Nations General Assembly, “General Assembly Appoints Artificial Intelligence Panel, Names Joint Inspection Unit Members, Notes Article 19 Arrears,” Meetings Coverage and Press Releases, GA/12751, 12 February 2026. Available at: https://press.un.org/en/2026/ga12751.doc.htm

[^3]: Michael Kratsios, “Remarks at the Security Council’s Open Debate on Artificial Intelligence and International Peace and Security,” United States Mission to the United Nations, 24 September 2025. Available at: https://usun.usmission.gov/remarks-at-the-security-councils-open-debate-on-artificial-intelligence-and-international-peace-and-security/

[^4]: United Nations Secretary-General António Guterres, “Statement upon the appointment of the members of the Independent International Scientific Panel on Artificial Intelligence,” Press Release SG/SM/23016, 12 February 2026. Available at: https://press.un.org/en/2026/sgsm23016.doc.htm

[^6]: United Nations General Assembly, GA/12751, op. cit.

[^11]: Kratsios, “Remarks at the Security Council’s Open Debate,” op. cit.

[^15]: Michael T. Klare, “UN Addresses the Promise and Perils of AI,” Arms Control Today, November 2025. Available at: https://www.armscontrol.org/act/2025-11/news/un-addresses-promise-and-perils-ai

[^16]: Milton Mueller, “Threat Analysis of ITU’s WCIT (Part 1): Historical context,” Internet Governance Project, 24 May 2012. Available at: https://www.internetgovernance.org/2012/05/24/threat-analysis-of-itus-wcit-part-1-historical-context/

[^17]: “World Summit on the Information Society,” Wikipedia, accessed 20 March 2026. Available at: https://en.wikipedia.org/wiki/World_Summit_on_the_Information_Society

[^18]: “Working Group on Internet Governance,” Wikipedia, accessed 20 March 2026. Available at: https://en.wikipedia.org/wiki/Working_Group_on_Internet_Governance

[^19]: “Internet Governance Forum,” Wikipedia, accessed 20 March 2026. Available at: https://en.wikipedia.org/wiki/Internet_Governance_Forum

[^21]: Mueller, “Threat Analysis of ITU’s WCIT,” op. cit.

[^22]: David P. Fidler, “Internet Governance and International Law: The Controversy Concerning Revision of the International Telecommunication Regulations,” ASIL Insights, Vol. 17, Issue 6, 2013. Available at: https://www.asil.org/insights/volume/17/issue/6/internet-governance-and-international-law-controversy-concerning-revision

[^24]: “Michael Kratsios at India AI Impact Summit 2026: A Vision for Sovereign AI,” Boston Global Forum, February 2026. Available at: https://bostonglobalforum.org/news/michael-kratsios-at-india-ai-impact-summit-2026-a-vision-for-sovereign-ai/

[^26]: United States Mission to the United Nations, “Remarks on the Appointment of members of the Independent International Scientific Panel on Artificial Intelligence,” 13 February 2026. Available at: https://usun.usmission.gov/remarks-on-the-appointment-of-members-of-the-independent-international-scientific-panel-on-artificial-intelligence/

[^27]: “Michael Kratsios at India AI Impact Summit 2026,” op. cit.

[^29]: Ken Moritsugu and Kelvin Chan, “U.S. rejects international AI oversight at U.N. General Assembly,” NBC News, 27 September 2025. Available at: https://www.nbcnews.com/tech/tech-news/us-rejects-international-ai-oversight-un-general-assembly-rcna233478

[^30]: Jared Perlo, “What the UN Global Dialogue on AI Governance Reveals About Global Power Shifts,” Center for Strategic and International Studies (CSIS), 14 October 2025. Available at: https://www.csis.org/analysis/what-un-global-dialogue-ai-governance-reveals-about-global-power-shifts

Methodology Note

This assessment is based on publicly available sources including official UN documentation, US government statements, academic analysis, and verified press reporting. All factual claims are traceable to primary sources or authoritative secondary sources.

ISAR Global’s analysis focuses on governance process intelligence—tracking what international coordination mechanisms actually achieve versus what they promise. This brief demonstrates systematic institutional analysis enabling accurate assessment of governance outcomes before they occur.

ISAR Global — Understanding governance reality rather than governance rhetoric

Contact: intelligence@isar-global.org
Web: isar-global.org

This intelligence brief is provided for informational purposes. ISAR Global maintains political neutrality and does not advocate for specific policy positions.