The Politics of AI Benefit-Sharing

Artificial intelligence technologies are developed, deployed, and governed within existing structural imbalances. Geopolitical rivalry, particularly between the United States and China as frontier AI developers, significantly influences how AI's benefits are diffused globally. International resource distribution can occur passively through market dynamics or proactively through mechanisms like benefit-sharing (e.g., the Windfall Clause). Here, we focus on the latter—its politics, challenges, and alternatives.

One of the key issues of international AI governance is how transformative AI systems may entrench or alleviate global inequalities. Given the current geopolitical dominance of the US and China, backed by their leading AI capabilities, AI benefit-sharing is most likely to continue on a trajectory that further reinforces global power concentration rather than change structures in favor of the Global Majority actors. We propose potential research directions to recalibrate the course of AI benefit-sharing towards more equitable scenarios.

About AI benefit-sharing

Transformative AI (TAI), as defined in a recent paper for the US National Bureau of Economic Research, refers to AI technologies that are capable of producing a societal transition at the scale of the next industrial revolution. For example, the advent of TAI may rapidly lead to skyrocketing market outputs and plummeting wages.

Additionally, automation of research and development may lead to significant acceleration in the rate of scientific discovery and innovation. Without adequate preparedness, such technological and economic progress density would likely result in highly asymmetric gains and broader societal chaos. That is because an intelligence explosion could lead to loss of control over TAI or TAI-enabled power concentration, among others.

However, few countries are ready to adopt the most advanced systems. This means that AI’s promises are likely to disproportionately benefit a very narrow part of the global population. Even though the risk of widening inequality due to the varied levels of AI adoption readiness constitutes both a domestic and an international challenge, we focus on the latter, as the domestic challenge is more likely to be addressed thanks to stronger incentive structures than at the international level.

International AI benefit-sharing refers to the ​​efforts to support and accelerate international access to AI’s economic or broader societal benefits. Yet, in reality, the very notion of “sharing” as well as the exact definition of “benefits” remain rather nebulous and contested: the former may mirror dependency, and the latter appears to deny the extractive roots of these profits. What is more, the exact degree to which AI benefit-sharing will rebalance global power structures and increase the Global Majority actors’ sovereignty or control over emerging technologies remains a tension between competing political agendas.

Historically, international efforts on benefit-sharing have had mixed results, rarely mitigating power imbalances. For example, in the context of global health, benefit-sharing tends to fall short of advancing equity. Despite a number of international agreements, the Global Majority countries’ access to sovereign genetic resources usually fails to be exchanged for the benefits associated with their use in scientific research conducted in the Global North, such as pandemic-related products. Similarly, despite the United Nations' altruistic call for the exploration of outer space to benefit all humankind, the Global Majority nations remain left behind in seizing space technologies to advance their economies and are subject to the major powers' politics.

Today, AI benefit-sharing discussions follow a similar trajectory toward reinforcing global power concentration instead of mitigating it. At the core, there is tension over what different actors perceive as the ultimate goal of these efforts. Global Majority nations emphasize asserting control over AI for sustainable development and ensuring sovereignty as AI economies, among others. However, even though equitable distribution that empowers marginalized communities at the international level is a praiseworthy aim from a global justice point of view, it barely influences decision-making.

Especially since AI benefit-sharing is increasingly enmeshed in US-China AI competition, Global Majority actors experience amplified tensions as they choose their allies and suppliers across the two sides. In this sense, AI benefit-sharing can be viewed as another form of soft power rivalry, like foreign aid, albeit possibly more consequential on the global order, given the weaker role of middle powers supplying AI resources than foreign aid. A more realism-driven logic can lead AI benefit-sharing advocates to approach the topic from a US-aligned perspective, that is, scenarios where the US uses AI benefit-sharing to reinforce its hegemony. While such a strategic approach may move the needle in the near term, it risks further concentrating power along the US-China nexus and failing to substantially improve Global Majority countries’ AI access and capacity to develop autonomy or leverage on the global stage.

There appears to be a neglect of a more global approach to AI benefit-sharing coordination amidst the current power dynamics. This is likely to cause both the Global Majority’s exclusion from AI benefits in the short term and from international AI governance in the long term. The challenge lies in building mechanisms for equitable AI access that are not overreliant on the control of the US or China, even though it is their AI development capabilities that dictate today’s global politics, and instead empower Global Majority actors.

As the nascent field of AI benefit-sharing takes shape, we see fostering discourse and redistribution scenarios that empower the Global Majority as key. Moving the needle on both requires alternative and more multipolar approaches that break the unpopular historical pattern of the Global Majority’s actual (dis)advantage from international benefit-sharing. As such, what could a coordinated effort of Global Majority states look like to claim AI benefits, deploy them, and magnify their profits? What mechanisms can be put in place, and by whom, in the next 6-18 months, to concretize global redistribution mechanisms before TAI’s disruptive effects?

Challenges

As AI’s economic integration remains limited for the time being, substantial benefits haven't fully emerged yet. However, four structural factors already play a critical role in preventing a more equitable distribution of AI promises:

• AI is a national security concern: Its dual-use nature—civilian and military—positions it strategically, causing governments to tightly control dissemination and significantly restrict international sharing.
• An absence of shared normative frameworks: Without shared international norms or standards around transparency and intellectual property management for AI access and deployment, uncertainty and mistrust inhibit collaborative benefit sharing.
• Private sector leadership and market constraints: Private sector actors dominate AI’s frontier development. Current market-driven incentives discourage investment in and diffusion to the Global Majority due to perceived profitability and regulatory risks.
• Development bottlenecks in the Global Majority: Educational gaps, inadequate infrastructure, technical limitations, and regulatory uncertainties further hinder AI’s global diffusion, particularly to the Global Majority.

Alternative scenarios

Given the existing challenges, we highlight two key directions for further research on AI benefit-sharing that empowers Global Majority actors:

Identify champions of equitable AI resource distribution: these champions require the ability to make demands at the global governance table. This could derive from factors such as their role in the AI supply chain, combined with market leverage. For example, what role could India play in pioneering AI benefit-sharing deals that empower Global Majority nations rather than fostering dependencies? It is the largest open market for tech giants globally, the second largest market for OpenAI, and a key growth market for Google. It will also host the next AI Summit later in 2025. India may hold promising market leverage to demand AI benefit-sharing scenarios on its own terms, which may include asserting access to both frontier capabilities and the necessary enablers for developing these frontier AI systems locally. More research is necessary to map out the points of bargaining power that Global Majority states possess to enable their more meaningful engagement with the leading AI actors. It also appears crucial to understand the exact potential of regional alliances and the role of middle powers in re-conceptualizing AI benefit-sharing as a tool for empowerment, mitigating global AI power concentration.

Outline the best-case scenarios for AI benefit-sharing: to develop the most suitable design for AI benefit-sharing, it is essential to first theorize its desired impacts. Some key knowledge gaps involve:

• What are the goals that AI benefit-sharing policies should be directed at? And what are the principles that should underlie AI benefit-sharing deals?
• How do policymakers mitigate the tensions between benefit-sharing and AI risks? What role can diplomacy and technical solutions play?
• Under what market and regulatory conditions would private frontier AI labs be incentivized to ethically share AI capabilities with Global Majority actors?
• What structural and institutional conditions must be met for Global Majority countries to effectively absorb and integrate advanced AI capabilities into their economic systems?
• What capacity-building initiatives (technical, educational, regulatory) are most effective for enabling Global Majority countries to sustainably deploy and manage transformative AI technologies?

Summing up, if AI can compress the rate of innovation discovery, there is an urgent need for a clearer vision of (1) a globally beneficial state of the world that AI benefits can be shared towards, (2) specific mechanisms to globally distribute benefits, decision-making power or independence, among others, (3) hierarchy of (international) policy and funding priorities. Just like the Copenhagen Consensus Expert Panel helped prioritize proposals for confronting ten great global challenges in 2004, what the nascent field of AI benefit-sharing needs today is a consensus-based identification of the most significant current challenges and a prioritized list of opportunities to address them.