PARIS: It is now abundantly clear that governments must take responsibility for shaping the AI industry. Only independent and proactive institutions, including regulators, can reduce the risks posed by these technologies while ensuring that they enhance, rather than undermine, human dignity.
Too often, however, governments have responded to the rise of AI with resignation, effectively admitting that they do not understand these systems well enough to intervene. In any case, many policymakers argue, it is too soon to justify meaningful action.
That is not how societies have managed previous technological revolutions. From ozone depletion to geoengineering and genetic modification, governments have intervened when there was credible evidence of harm, or when the potential for serious harm became too great to ignore, rather than waiting for complete scientific certainty. In doing so, they have generally adopted an outcome-based approach, focusing on the real-world impact on people and holding companies accountable for the harms caused by their products or technologies.
Governments have a duty to require rigorous assessment of powerful new technologies before they are rushed to market. Independent agencies such as the US Food and Drug Administration, the European Medicines Agency, and the European Food Safety Authority already perform this function for food and pharmaceuticals.
No one, of course, expects government agencies to invent better drugs than pharmaceutical companies can. Their role is to verify that the medicines those companies produce are safe and effective.
And despite decades of warnings that regulation would stifle innovation, independent oversight has not made the industry any less innovative or competitive. On the contrary, it has made its discoveries more trustworthy.
Yet the US government, whose decisions shape the global AI market more than those of any other, has repeatedly balked at regulating these technologies, invoking the familiar argument that regulation kills innovation. This reluctance has been reinforced by an increasingly zero-sum view of technological competition with China, even though the United States still develops the overwhelming majority of frontier AI models while China accounts for only a small fraction.
In keeping with the neoliberal approach that has dominated economic policymaking for the past half-century, governments have generally been expected to intervene only after markets fail. But when it comes to AI, a wait-and-see approach could prove catastrophic. Anthropic’s testing of Mythos, for example, showed that the model could identify and exploit vulnerabilities across all major operating systems and web browsers. If deployed irresponsibly or placed in the wrong hands, frontier AI systems with such capabilities could compromise critical infrastructure, financial systems, and national security on a global scale.
In response, US President Donald Trump’s administration has recently imposed restrictions on certain foreign actors’ access to frontier AI models, encouraging countries around the world to reduce dependence on American AI providers and invest in domestic capabilities. This is not the kind of intervention that democratic societies should welcome.
Trump took this logic a step further in June by suggesting that the federal government acquire equity stakes in AI companies so the “American people can benefit.” Just weeks later, OpenAI reportedly proposed offering the US government a 5% stake in the company.
But government ownership should not be confused with public ownership. A public good is something everyone can use and from which no one can be excluded, such as clean air, a public park, or a lighthouse. An equity stake, by contrast, confers ownership rights and influence over a private company, exercised by whichever government happens to be in power.
The OpenAI proposal raises profound concerns. The first is privacy. Imagine turning to an AI assistant like Claude Fable with the questions you would never ask anyone else: how to leave a marriage, how to hide debt from your family, or whom to vote for. Increasingly, people treat AI systems like diaries that answer back.
Now imagine that the government owns part of the company operating that system. In 1984, George Orwell imagined a tyrannical state that imposed surveillance through telescreens. AI offers a subtler dystopia, one in which surveillance begins not with coercion but with trust.
The second concern is institutional. Governments exist to ensure that markets benefit society, not to maximize the value of their investment portfolios. If the state becomes a shareholder in major AI firms, those objectives may come into conflict. Would policymakers be willing to sacrifice the value of the government’s investment in the name of long-term public safety? Even if governments acted in good faith, the structural conflict of interest—or even the appearance of one—would undermine confidence in regulatory independence.
Moreover, OpenAI’s proposal did not emerge from careful institutional design or democratic deliberation. It appears to be an ad hoc response to immediate political pressures rather than part of a coherent strategy for governing AI.
Decades of experience have also taught us what effective intervention looks like: independent regulators, impartial institutions, protections against conflicts of interest, and strict limits on revolving doors between the public sector and industry. Those lessons should not be abandoned precisely when they are needed most.
There are far better ways for policymakers to steer AI markets in the right direction than taking ownership stakes. Rather than becoming shareholders, governments should invest in strong, independent institutions, develop and enforce safety standards, and establish clear liability regimes so that those responsible for AI-related harms can be held accountable. They should also require multinational tech companies to pay their fair share of taxes. Above all, they must cultivate the public-sector expertise needed to keep pace with technological advances.
These proposals are far from radical. They are the institutional tools that successful market economies have developed over decades to reconcile innovation with responsible governance. As transformative as it may be, AI does not require us to reinvent the role of government. But we must reclaim that role. Rather than becoming shareholders, governments should do what they do best: shape markets in the public interest.
Gabriela Ramos, Co-Chair of the Task Force on Inequalities and Social-Related Financial Disclosures, is a former assistant director-general for social and human sciences at UNESCO, where she oversaw the development of the Recommendation on the Ethics of AI, and a former OECD chief of staff and sherpa to the G20, G7, and APEC. Emilija Stojmenova Duh, Associate Professor of Electrical Engineering at the University of Ljubljana, is a member of the Globethics Foundation Board, a member of the EU AI Scientific Panel, and a former minister of digital transformation of Slovenia.
Copyright: Project Syndicate, 2026.
www.project-syndicate.org
