Our AI Lawless Era: What the Thirteen Years Before the First Driving School Teach Us About Governing Artificial Intelligence

Chapter proposal submitted to the edited volume "Artificial Intelligence, Governance, and Sustainable Development: Ethics, Policy, and Global Challenges" (Springer Nature), November 2026. In 1886, Karl Benz patented the automobile. It was not until 1899 that anyone opened a school to teach people how to drive one. In the thirteen years between, the road became a laboratory of harm: the first pedestrian deaths, the first liability insurance policy, speed limits pegged to the pace of a horse, and men walking ahead of cars waving red flags. Authorities restricted the machine, punished drivers and warned the public, yet casualties kept climbing. What finally changed the trajectory was not a better engine but a recognition that had been missing all along: a powerful technology is only as safe as the people trained to operate it. This chapter argues that we are now living through artificial intelligence's equivalent of those thirteen years. Through a historical comparative analysis of the early automobile era (1886–1913) and contemporary evidence on AI adoption, incidents and governance, it identifies a recurring pattern in which transformative technologies are deployed under competitive pressure long before the educational and institutional infrastructure needed to use them safely exists. Today, many organizations adopt AI chiefly because their competitors do; reported AI incidents have multiplied since 2022; only a small minority of organizations have a functioning governance framework, even as most express confidence in their risk management; and large shares of employees use AI tools without disclosure or oversight. Formal regulation, led by the EU AI Act, is arriving, but its phased timelines trail deployment, much as speed limits and licensing once trailed the car. The chapter makes three contributions. First, it proposes a "pre-driving-school" framework for diagnosing governance lag. It maps the four responses that failed in the automobile era (restricting the technology, mandating warnings, escalating penalties and running awareness campaigns) onto their AI counterparts, showing how each treats symptoms rather than the root cause: untrained operators. Second, it reframes AI literacy and operator training not as a corporate nice-to-have but as core governance infrastructure, the missing layer between high-level ethical principles and everyday use, without which laws and standards remain difficult to implement. Third, it examines where the analogy breaks down. Unlike a runaway car in 1895, a poorly governed AI system can harm millions simultaneously, cross borders instantly and fail invisibly. Society cannot afford a thirteen-year learning curve, nor can it afford to learn only from accumulated casualties. These arguments are then connected to the volume's central question: how AI can be governed so that it advances rather than undermines the Sustainable Development Goals. Untrained adoption carries real economic cost — failed projects, data leaks, reputational damage and eroding institutional trust — which bears directly on decent work and economic growth (SDG 8). Those costs fall unevenly: institutions and countries that can afford governance expertise, legal review and in-house technical capacity are able to absorb AI's risks, while those that cannot bear them in full, widening the very inequalities SDG 10 seeks to narrow. Weak oversight of AI in public administration likewise undercuts the accountable and transparent institutions envisioned in SDG 16. The chapter's central claim here is that SDG 4's commitment to inclusive, lifelong learning offers the most immediately available lever: unlike regulatory capacity, which takes years to build, AI literacy can be delivered now, at scale, and to the people already using these systems. Particular attention is given to the Global South, where AI tools are arriving well ahead of local regulation and where workforce training may be the most practical first line of governance. The chapter concludes with recommendations for governments, educators and organizations: establishing baseline AI competency, analogous to a driving licence, for high-risk uses; embedding role-based training in national AI strategies and corporate governance; building shared incident-reporting mechanisms so that harms inform policy before they compound; and designing the "driving schools" of the AI age as public goods rather than privileges. It closes with the question history poses plainly: must we accumulate our own decade of AI accidents before we teach people to drive, or can we, this time, build the school before the crash? Includes the original essay "Our AI Lawless Era" (first published February 2026), from which this chapter proposal developed.