If AI is a public good, nationalisation should be considered

ARTIFICIAL intelligence could well prove to be the most profoundly consequential technology in human history. It is already challenging long-held notions of intelligence and creativity, promising to hasten scientific progress, and raising uncertainties about the future of jobs.

In these tantalising times, there is no shortage of techno-optimists who insist that we stand on the cusp of a world of super-abundance as AI usage will enable historic gains in productivity and prosperity.

Many economists, however, are not quite as sanguine. We worry about the gains being concentrated among a few, while large swaths of white-collar jobs are automated away.

The spectre of widespread unemployment and a steep rise in inequality worries us, with clear consequences for social stability. We think about policy solutions such as taxing AI companies and using the proceeds to fund large-scale public-sector transfers to the general population to mitigate such risks.

There is a third camp of AI observers that goes even further, claiming that unless regulated stringently, forthcoming AI models and applications, so potentially powerful and destructive if misused, could prove to be catastrophic. From individual data privacy breaches to enterprise security collapse, global financial instability to outright weaponisation, the scenarios are deeply unsettling.

Governments and companies talk about ethical standards and AI boards, global cooperation and regulation, but we are surprised to find so little attention given to a solution that seems obvious – nationalisation, or at the very least, having the public sector take a substantial equity stake in major AI companies.

We offer a few reasons why this might be a good idea.

First, safety. There is increasing evidence that AI can be trained to be a super hacker, undermining conventional guardrails around data and associated cyber security measures. Financial systems could be disrupted and financial wealth stolen or destroyed. Public utilities, from transportation to power to water supply could be cut off. A rogue actor could use it to design a maximally lethal virus.

Consider an analogy with nuclear technology, which has dual use applications. When harnessed for peaceful purposes, it is a reliable source of clean energy.

But it can also be used to develop weapons that have the power to destroy humanity. We don’t let the private sector have access to nuclear weapons; in fact, all nuclear technology is under state control and bound by global non-proliferation treaties.

While nuclear power can be in the hands of the private sector, it is regulated through a combination of strict international standards, licensing process, and national laws, with rigorous oversight in every phase of a facility’s lifecycle. Should the same be done for AI?

One criticism of this line of reasoning would be that it is too late. AI is general purpose and already diffused across users, developers, and open-source models.

But just like central banking came long after money began circulating and utilities came long after electricity and water supply, the widespread rollout of the technology cannot be an excuse for giving up on regulation and vigilant oversight.

Second, strategic importance. Governments have traditionally sought to keep the commanding heights of the economy in public control, and AI’s strategic importance is of the first order. In his seminal work, Capital in the Twenty-First Century, Thomas Piketty has calculated that public capital accounted for 20 to 30 per cent of total capital in Western democracies in the 1970s.

There were good reasons for this. But these have been largely forgotten in the privatisation wave that has swept the world since the 1980s. Among high-income economies, Singapore is a notable exception, with the public sector owning the bulk of the land and a large part of the capital stock of the economy.

Given the strategic importance and sensitivities associated with AI, a case can be made for deeper public sector involvement through equity ownership. Options include setting up a publicly funded AI competitor, golden shares in AI companies held by sovereign wealth funds, and multilateral treaties to standardise oversight and information sharing.

Perhaps all of these options should be deployed simultaneously and expeditiously, given the pace of ongoing developments.

There is an additional argument for public equity in AI. The technology behind AI arises in large part from decades of publicly funded research, suggesting that the public should gain from the upside. This point has been made forcefully by economists Mariana Mazzucato and Tommaso Valletti.

Third, distributional fairness. AI, by design, is extremely capital-intensive and labour-light. It therefore seems destined to raise the capital share substantially and shrink the labour share of national income. Fairness and social stability demand that ordinary citizens benefit from the upside through a capital dividend attached to public ownership. This would be more socially palatable than universal basic income.

Finally, common imperative. Public control would reduce the risk of concentration of power among the few owners of dominant AI companies. Technology oligarchs can potentially use their enormous resources to override the preferences of the democratic majority, a risk that must be recognised.

Concern for the common good, especially regarding potential AI overreach in society, cannot be left to the whims of technology billionaires. Additionally, AI models are trained on content generated by all of us, suggesting that this technology should be regarded as part of the commons.

As for social choice, economists Daron Acemoglu, David Autor, and Simon Johnson have argued that AI could be used to either augment or replace labour. This reflects a choice. The government would have greater incentives to tilt AI in a pro-worker direction, instead of simply accepting widespread automation, which would lead to massive social fallout.

Above all, something so consequential, potentially disruptive and paradigm-shifting cannot be managed just through basic regulation. A far more deliberate intervention is required to monitor, develop, safeguard and tax this technology. The public interest could not be greater.

Anthony Annett is an author and former staff at the International Monetary Fund; Taimur Baig is managing director and chief economist at DBS Bank