Technological sovereignty alone will give AI a gilded age

The session on the future of artificial intelligence asked how AI should be governed in a world of competing powers. I answered with a prior question: what role is the state now taking in technological change, and is it the role this stage of the digital revolution requires? The argument comes from a CEBRI policy paper I wrote in 2020; the closing part, on the governance of AI, draws on a later essay for CIGI.

Karl Polanyi described capitalist development as a double movement: each expansion of the market meets a countermovement of social self-protection. Read as a pendulum, it gives a rough calendar. Free-market periods ran from 1830 to 1870, 1920 to 1940 and 1970 to 2010; the latest swing towards state-centred self-protection began after the financial crisis of 2007–08. The pandemic accelerated it. Value chains dependent on China became a target of national policy, and industrial policy, increasingly framed as mission-oriented, returned to the menu. The swing is clearest in discourse: government documents in the Overton database citing technological sovereignty were almost absent before 2019 and exceeded three hundred in 2021.¹

Carlota Perez’s theory of technological revolutions explains why the pendulum moves. The fifth revolution, opened by the microprocessor in 1971, was installed under financial capital and free markets. Its deployment calls for an active state that supports production capital, regulates the new technologies’ externalities and addresses the inequalities their installation produced. The fourth revolution had its framework: the New Deal, Bretton Woods, Galbraith’s countervailing powers, and successive safety, consumer and environmental laws.

The present return of the state lacks that framework. Its rhetoric treats governance and regulation as instruments of technological sovereignty. AI raises ethical, proprietary, industrial-design, normative, techno-economic and socio-environmental problems, and several of them cannot be settled by national rules alone. Geopolitical competition, meanwhile, narrows the space for multilateral governance, and what it tends to produce is a contest of standards: American, Chinese and European. Competing standards slow development, impair diffusion and raise costs, at a time when environmentally sound AI solutions need to diffuse fast.

Perez sees a golden age as a possibility in the second half of every revolution, provided institutions address the problems created in the first. The post-war golden age emerged during a cold war, so geopolitical rivalry does not rule one out. Whether this swing of the pendulum produces one depends on states going beyond a narrow definition of sovereignty and building, together, the institutions that the deployment of AI requires. Otherwise, the closer precedent is the gilded age of the third revolution.