Skip to content
Technology and Innovation

CFR Surveyed 350 Experts About AI’s Future. Most Think Governance Is Failing.

CFR asked experts to predict how artificial intelligence (AI) and governance will affect the global order in 2035. Almost all agreed that governments are failing to keep pace with AI, the future of frontier AI capabilities is highly uncertain, and AI labs are already gaining power relative to the public sector.

Scatterplot on a 2x2 axis
Photo illustration by CFR

By experts and staff

Published
  • Jessica BrandtCFR Expert
    Senior Fellow for Technology and National Security
  • Adam SegalCFR Expert
    Ira A. Lipman Chair in Emerging Technologies and National Security and Director of the Digital and Cyberspace Policy Program

The decisions being made about artificial intelligence (AI) right now—who builds it, who governs it, and who benefits from it—will shape geopolitics for decades to come. CFR surveyed three hundred fifty foreign policy experts about what they expect the distribution of frontier AI capabilities and the nature of governance structures that emerge around them to look like by 2035. The experts disagreed on almost everything, but what they did agree on was alarming: no one is prepared to govern the AI capabilities that could be coming.

Read the main takeaways below and the full survey results on CFR.org.

Experts Agree That Governance Is Failing

More than 80 percent of respondents expect that AI governance will be fragmented. Only 5 percent anticipate that domestic institutions, such as regulatory bodies and courts, will keep up with the pace of AI development.

A U.S.-China bilateral agreement is the path experts see as the most likely option to produce a binding governance agreement—a dim prospect, given the state of U.S.-China relations, and one that would leave most of the world on the sidelines.

Meanwhile, more than 70 percent of respondents say a serious AI accident is the most likely trigger for change. Taken together, those findings suggest that things will get worse before they start to get better.

Experts Are Genuinely (and Revealingly) Split on Who Will Build Frontier AI

Respondents came to no clear consensus on how concentrated frontier AI will be in 2035. About half (46 percent) expect the ability to develop frontier AI capabilities to remain in the hands of a few actors; 54 percent predict it will diffuse to dozens of states and nonstate actors. Few experts land between those options.

That polarization likely reflects two incompatible views on how AI will develop. One group expects scaling laws to hold, and that increases in model size, data, and computing power will reliably translate to improvements in AI performance, keeping frontier AI development costly and, therefore, concentrated. The other expects that new models, algorithms, chips, or other technological breakthroughs will make frontier AI models smaller, cheaper, and more accessible to more developers.

The economic stakes of frontier AI development are considerable. Most experts (68 percent) expect productivity gains from AI to concentrate among advanced economies or a small number of private actors. Experts are skeptical that new alliances or political strategy will change that calculus for the Global South: 64 percent say receiving meaningful benefits from AI would require a fundamental shift in how AI is built—one that moves away from models demanding enormous data, energy, and compute, and toward alternatives that are less resource intensive.

Power Is Already Shifting From States to Labs

Almost 70 percent of respondents believe frontier AI labs will be the most powerful nonstate actors by 2035, and 75 percent expect nonstate actors to gain leverage over the state.

Those predictions are already coming true: at the Group of Seven Summit in June 2026, world leaders were joined by the CEOs of OpenAI, DeepMind, Anthropic, Mistral, and other companies to discuss AI governance, AI infrastructure, and international stability. Respondents would not have been surprised that the labs had a seat at the table, but they would have worried about what their presence means for governance. Along with labor market disruption and democratic backsliding, respondents identified regulatory capture as one of their top concerns, fearing AI labs will gain power over the governments that are meant to oversee them.

Where Things Stand

As the survey results highlight, experts broadly agree that the world is navigating profound uncertainty without agreed-upon rules or institutions that can ameliorate the social, political, economic, and strategic instability that will accompany rapid AI development and deployment. Experts are split over where AI is heading, but nearly unanimous that no one is steering. Frontier AI labs have the most power and expertise, but they, so far, remain unaccountable to governments or the publics they affect.

Subscribe to receive updates from this project on AI and the next global order in your inbox.

This work represents the views solely of the author(s). The Council on Foreign Relations is an independent, nonpartisan membership organization, think tank, and publisher, and takes no institutional positions on matters of policy.