What Should Governments Do in the Age of Artificial Intelligence—A Schumpeterian Growth Paradigm Perspective

GAICIL Insights

Governments keep asking how to regulate artificial intelligence.¹ Philippe Aghion's article on what governments should do in the age of artificial intelligence, published in Research in Government and Economics, asks a larger question: how to make an economy thrive through it. AI is the latest form of automation technology, spreading across production and services, and the article draws on growth theory built around creative destruction to assess what the AI revolution means for productivity, inequality and public policy.

The central prediction is optimistic. Like the electricity revolution and the information technology revolution before it, the AI revolution is expected to enhance corporate productivity and raise the potential for economic growth. Every update and application of AI in production, in services and in the production of ideas reshapes the industrial landscape and drives the reallocation of resources. The process closely fits the creative destruction described by innovation-based growth theory, and every industry will be affected. No sector gets to sit this one out.

The risks are named just as plainly. AI can promote growth, yet it may also produce a growth dilemma, exacerbate income inequality and educational inequality, and impose significant disruptions on nations' economic development and social fabric. The reallocation that raises productivity displaces existing activities and workers. Governments need systematic policy measures to respond to these shocks. The engine and the wreckage come from the same mechanism.

Industrial policy that encourages innovation and competition is the key to securing growth. Such policy must let firms enter, grow and earn innovation rents. It must also ensure that, once firms have grown, their expansion does not come at the expense of new firms' entry. That dual requirement is the central tension in AI markets: reward today's winners without letting them lock the door behind them.

Entry is the pressure point in AI. Leading AI developers benefit from scale in data, computing power and talent. If these advantages harden into barriers, new firms may struggle to enter, and the creative destruction that drives growth could slow. Competition policy protecting contestability supports the dynamism the article identifies as the source of growth. A frontier owned by three firms is a frontier that may stop moving.

Education is the second pillar. AI may widen educational inequality as well as income inequality, so governments should improve educational systems. Workers able to use and complement AI are likely to gain, while others risk being left behind. Education that builds the skills to work alongside AI spreads the productivity gains.

Social protection and redistribution form a third pillar. The article recommends strengthening social security frameworks and using tax policy to balance income distribution. Creative destruction creates winners and losers; social insurance cushions displaced workers and makes reallocation more acceptable, which sustains support for innovation. Insurance is what keeps the public from voting against the future.

Government itself must change. The article calls for an expanded role for government institutions and for models of social governance adapted to AI. Public administration must evolve alongside the technology, both to use AI effectively and to govern its effects.

International cooperation closes the agenda. Balanced AI development requires multilateral consultation among governments on AI-related infrastructure, data privacy, algorithmic bias and intellectual property. Because AI crosses borders, coordination can reduce fragmentation and help more countries benefit.

History gives the analysis its force. Previous general-purpose technologies raised productivity over time while disrupting existing arrangements. Placing AI in this lineage suggests its growth potential is real and that policy choices will shape how widely its benefits are shared.

On policy, the message is a balanced agenda: foster innovation and competition, improve education, strengthen social security, balance income distribution and promote multidimensional international cooperation. Each element addresses a different channel through which AI affects economies and societies.

Stepping back, the AI revolution can raise productivity and growth, much as electricity and information technology did, while creating disruption and inequality. Combining industrial policy that rewards innovation and competition with education, social protection and cooperation captures the gains while managing the costs.

For AI ministries, the sting is this: a strategy consisting only of safety rules and compute subsidies covers a fraction of the agenda. The countries that win may be the ones that also keep entry open, retrain workers and insure the losers.