AI regulation answers the wrong question

Attend almost any AI conference today and the conversation turns to regulation. Governments are drafting AI legislation (Opens in new window) while commentators warn of deepfakes and existential risk. It is easy to conclude that AI's central policy challenge is regulation.

History suggests we should begin with a different question.

What enduring functions is AI transforming?

To illustrate the point, go back to industrialisation and the transformation of Britain’s economy. Governments initially focused on protecting existing economic interests, such as via the Corn Laws (Opens in new window). Only gradually did attention shift towards institutions, including factory inspectorates and public health authorities.

Responding to the revolution of AI will likewise depend on two forms of innovation. The technological innovation creates new possibilities. Institutional innovation determines whether and when those possibilities translate into lasting prosperity.

Concern over the consequence for labour offers another example. With every major technological revolution, the demand for labour has changed. Workers acquire new skills, firms reorganise production and occupations evolve. This is an inherent feature of economic development rather than institutional failure.

The relevant policy question is not whether AI changes jobs, but whether it creates new collective action problems that existing labour market institutions cannot address. If not, the appropriate response is to strengthen education, training and labour market institutions rather than treat AI as a fundamentally new employment problem.

Here, institutions (Opens in new window) are understood broadly to include the rules, organisations, capabilities and coordinating mechanisms through which societies organise collective action. The central question is therefore not this week’s AI controversy, but which institutional arrangements will best support the public goods, mitigate the public bads and enable the markets an AI economy will require. If it will not still be needed 20 years from now, policymakers are probably responding to a transient technology rather than an enduring institutional requirement.

Institutional lags are inevitable because societies need time to discover the collective challenges technological change creates. The lesson of history is that governments need not prolong that lag by allowing reactive policymaking to divert attention from the institutions a new economy ultimately requires.

AI is an opportunity to shorten that lag. The challenge is to identify the institutions an AI-enabled economy needs – rather than be distracted by today’s AI headline.

Like electricity or the internet, AI is a general-purpose technology (Opens in new window) embedded across the economy. AI is reshaping how societies create knowledge, make decisions, organise production, coordinate complex systems, develop human capability and sustain trust. It is the wrong object of regulation. The more useful question is what new economic and social functions AI creates, and what institutional capabilities those functions require.

The starting point is diagnosis rather than regulation. Governments cannot predict every consequence of technological change, but they can identify the institutions a new economy requires by focusing on enduring governance challenges and identifying public goods, public bads and missing markets rather than reacting to successive controversies.

The implications become clearer when viewed through the lens of public goods and public bads. Trusted evaluation systems, technical standards and governments capable of understanding AI are public goods that benefit society but are unlikely to be supplied adequately through private incentives. Fraud, misinformation and cybercrime are public bads whose defining characteristic is that their costs extend beyond those responsible for creating them.

The Industrial Revolution again illustrates the point. Britain prospered not because it regulated steam engines, but because it developed institutions to address the collective action problems industrialisation created. Public health authorities responded to the health risks of rapidly growing cities. Company law evolved to mobilise capital for large-scale industrial enterprise. The technologies themselves were rarely the problem. They exposed governance questions that existing institutions had never been designed to answer.

The same framework also broadens the range of institutional responses available to governments. Current debates over personal data focus largely on privacy regulation. Another response is to define property rights over personal identity and data, allowing individuals to control, license or withhold their use through contracts. Australia’s data portability rights (Opens in new window) go in this direction. Creating markets can be as important an institutional innovation as regulation itself.

Some challenges can be addressed through existing institutions and legal frameworks; others require new standards, public capability, market design or entirely new institutions. Regulation is only one institutional response, and only a minority of issues are likely to require AI-specific regulation.

Societies prosper not because they regulate technological revolutions particularly well – but because they develop the institutions that enable technological revolutions to create lasting economic, human, institutional and social capital.