Who benefits from AI depends on how it is governed

The hype and the often-polarised narratives surrounding artificial intelligence (AI) can make it difficult to understand what is really happening. In all the noise, it is easy to overlook the institutions and political choices that influence how technological change unfolds and who benefits from it.

A recent Eurofound research paper, AI, algorithmic management and the transformation of society, work and employment – Exploring the concepts, places institutions at the centre of its research framework for understanding AI, highlighting how the rules and systems that organise economic and social life shape technological change.

The institutional landscape around AI is itself evolving. As AI companies develop increasingly powerful technologies, they are also taking decisions about how those technologies should be assessed, what risks are acceptable and when systems should be deployed. At the same time, governments are becoming increasingly involved, seeking to assess, constrain and shape the development and use of frontier AI.

The emerging relationship between companies and governments raises important questions about who is shaping the trajectory of AI and on what basis. Recent developments involving Anthropic and the US government offer a useful illustration.

Not long ago, Anthropic, the company behind Claude, warned about the risks of AI improving itselfopens in new tab, even suggesting that frontier AI development might need to slow down. Shortly afterwards, however, it released its most powerful model yet. And the story did not end there. Access to some of its models was later restricted following a US government order over alleged security concerns, amid broader disputes over the conditions and safeguards governing the use of advanced AI.

Anthropic is not alone in deciding for itself how the risks of its technology should be managed. OpenAI, for example, recently paused work on an experimental model after it demonstrated advanced autonomous hacking capabilities in the now infamous OpenAI–Hugging Face incident.

At the same time, governments are becoming more directly involved in these decisions. In June 2026, the Trump administration issued an executive order establishing a framework for government assessment of advanced AI models, including their cybersecurity capabilities.

These developments complicate the familiar opposition between regulation and innovation. AI companies are increasingly making decisions about the risks their technologies pose and the conditions under which they should be developed and deployed, while governments are seeking a more direct role in assessing and shaping those decisions.

This should serve as a wake-up call for European policymakers to ensure that they have a say in how AI is being governed, rather than leaving those decisions to companies and governments beyond Europe’s control.

As the boundaries between technological development and governance are becoming increasingly blurred, we need to determine who shapes AI’s rules and priorities, who controls access to it and who decides how it is used. History shows that institutions have an important role to play in shaping how societies respond to profound technological change.

Eurofound’s approach to AI research draws on the work of economist Carlota Perez in treating technological revolutions as not just waves of innovation, but as profound economic and social transformations. Over time, they give rise to new techno-economic and socio-institutional paradigms. These transitions are rarely smooth or painless; their long-term success depends on the ability of institutions to adapt and to create the conditions for new technologies to spread across the economy and society.

If the pace of AI development feels unprecedented, history can still help us make sense of what is happening today and remind us of the decisive role of institutions. The Industrial Revolution, which began in Europe, brought extraordinary technological progress but also profound social disruption. Despite rapid productivity growth, the benefits of early industrialisation were slow to reach ordinary workers, with real wages lagging behind productivity growth for several decades.

Shared prosperity did not come from new machines alone; it emerged as societies built the institutions needed to channel technological change towards social progress. Labour laws, public education systems, social insurance and social dialogue did not emerge despite industrialisation, but because of it. The AI age may require a similar effort. But what should that effort seek to achieve?

The discussion of the role of institutions raises perhaps the most important question of all: as AI develops at extraordinary speed, will our institutions be capable of steering technological change towards broader social progress?

Or perhaps we should first ask a more fundamental question: what kind of society do we want AI to help create, and what kind of institutions are needed to make that possible?

Answering these questions means making collective choices that cannot be left to technology companies or market forces alone. The stakes could hardly be higher. The implications of AI extend far beyond technology itself, touching on security, democracy, social cohesion and the future of work.

If institutions shape technological trajectories, then the challenge is not simply to regulate new technologies. It is to ensure that innovation is guided towards widely shared social goals, and that productivity gains translate into better jobs and living conditions.