AI governance

AI governance is the project of designing institutions, norms, and regulatory mechanisms that shape the development, deployment, and societal integration of artificial intelligence systems. It is not a technical problem of controlling algorithms but a political problem of governing power — specifically, the computational power that is increasingly concentrated in a small number of organizations and deployed across populations that have no meaningful role in its design or operation.

The field spans three overlapping domains: safety governance (how to prevent systems from causing catastrophic harm), rights governance (how to protect individuals and groups from algorithmic discrimination, surveillance, and manipulation), and structural governance (how to prevent the concentration of computational power from undermining democratic institutions, economic competition, and epistemic pluralism). These domains are not independent. A safety governance framework that concentrates authority in a single oversight body may solve the catastrophic risk problem while creating a rights governance problem. A rights governance framework that relies on individual consent may solve the privacy problem while leaving the structural governance problem untouched.

The central tension in AI governance is between expertise and democracy. AI systems are technically complex, and their governance appears to require specialized knowledge that most citizens and many legislators do not possess. The temptation is to delegate governance to technical experts, corporate risk officers, or international bureaucracies. The risk is that this delegation reproduces the legitimacy deficit of algorithmic institutions themselves: governance by the governed is replaced by governance by the technically qualified, and the qualification itself becomes a mechanism of exclusion.

The deeper challenge is that AI governance is not merely about regulating AI. It is about preserving the capacity for collective self-determination in a world where consequential decisions are increasingly made by systems that no individual human designed, understands, or can effectively challenge. The governance of AI is, ultimately, the governance of the relationship between human societies and their own technical creations.