“Industrial policy” has renewed currency in several parts of the world today (see here and here). The term is being used to describe an uptick in state-led investment and policy strategies geared towards promoting particular sectors of the economy. In the U.S, there is active debate around several facets of this return to industrial policy: is it simply regurgitating a familiar playbook which sees governments adopting deregulatory postures despite advancing policy that cultivates domestic monopolies (“national champions”)? Or is there still the promise of a politics that supports an interventionist role for the state to direct the economy in line with public values, which could help undergird a “Green New Deal”-style approach?
Another open question is the extent to which these developments are motivated by the national security imperatives and backdrop of the “great power conflict” with China. In the European Union, which has been committed to “market efficiency” logic as a justification for EU’s expanding power, there appears to be greater willingness for state intervention in the market — from decarbonization policy to outlining a “digital sovereignty” vision through the EU Chips Act.
AI and semiconductors are a key sectoral focus for this new industrial policy agenda in multiple jurisdictions – from the CHIPS Acts in the US to equivalent legislation in the EU, France’s construction of the Jean Zay supercomputer and the UK’s $900mn commitment to similar efforts, to the Indian government’s claim to leading the “digital sovereignty” agenda for the Global South. With AI being heralded as the digital infrastructure of the future, we are seeing a wave of countries articulate their place in it, as they scramble to deploy a range of investment, policy, and narrative strategies to establish leadership on AI.
In contrast to industrial policy developments around green technology spearheaded by a broad progressive coalition (including organized labor), the current trajectory of AI industrial policy is being developed largely in closed quarters by technocratic elites – and is narrowly focused on leveraging private capital towards national security goals (especially via the so-called US-China AI Arms Race), or cultivating domestic monopolies (“national champions”). Within this paradigm, there is relatively little attention to challenging concentrated power in the tech sector; institutionalizing accountability frameworks that would temper the speed and scale of AI proliferation; or de-escalating investments in AI due to environmental consequences.
Building on our previous work on AI/data policy as industrial policy (see here, here and here), and in collaboration with a network of researchers and advocates, AI Now is curating a compendium to analyze these national developments on industrial policy for AI within their domestic and historical contexts. The category of “industrial policy” is interpreted dynamically to include not just direct subsidies, investments, or export controls but equally, policy approaches like antitrust, IP, and algorithmic accountability regulation that are used to shape industry outcomes as well as soft power narratives around AI leadership.
It’s early days for this space, evidenced by the fact that across regions we are seeing more noise than action (and relatively low levels of investment). Equally, it’s the ideal time to intervene to ask hard questions of the narrow motivations that currently animate this agenda, as well as of the often undemocratic modes by which it is being developed. This compendium aims to lay the foundation for civil society to redirect or subvert this limited imagination in favor of a more democratically contested industrial policy. In keeping with the notion that there is nothing inevitable about AI development, such a move can leverage the future trajectory of the tech industry against the backdrop of rising inequality, growing authoritarianism, and the ever more urgent climate crisis.
We look forward to sharing more from this work as it moves forward, and welcome insights along the way.