The Beginner's Guide to AI Governance

This seven-part beginner’s guide explains how AI governance is made and applied, from the reasons AI poses a policy challenge to the EU AI Act, national approaches in China, the UK and the US, and international cooperation. It draws on the author’s notes from an LSE programme on AI law and policy, supplemented with developments after the course material was prepared. Its central premise is that governing AI requires a workable shared definition: the term covers computational systems that learn from data to analyse, predict or recommend, not just chatbots or human-like machines.

The opening lesson argues that AI is a general-purpose technology whose reach across industries, continuing improvement and capacity to enable new products are matched by unusually rapid adoption, since it uses existing digital infrastructure. Its effects are uneven: it can automate some cognitive tasks while complementing others, but benefits depend on training and redesigned workflows, and may come with job displacement, workplace surveillance and reduced worker autonomy. The guide also identifies regulatory competition between jurisdictions, security risks involving misjudgement, escalation and dangerous knowledge, and the pressures of US-China rivalry. Research on productivity and employment offers conflicting results: AI has improved some workers’ speed and output, helped less-experienced customer-support staff more than veterans, and reduced earnings for some freelancers. Because forecasts about growth, jobs, elections and advanced capabilities diverge sharply, the guide presents uncertainty itself as a central challenge for policymakers.